release: OpenCV 3.4.17

Merge pull request #21341 from alalek:issue_20930_3.4
python: use '((x,y), (w,h), angle)' in std::vector<RotatedRect>
2026-07-21 19:33:03 +04:00 · 2021-12-24 16:39:15 +00:00 · 2021-12-24 16:33:19 +00:00 · 2021-12-24 15:01:45 +00:00 · 2021-12-24 11:56:12 +00:00 · 2021-12-24 11:55:55 +00:00
717 changed files with 39765 additions and 13280 deletions
@@ -1,5 +1,5 @@
 <!--
-If you have a question rather than reporting a bug please go to http://answers.opencv.org where you get much faster responses.
+If you have a question rather than reporting a bug please go to https://forum.opencv.org where you get much faster responses.
 If you need further assistance please read [How To Contribute](https://github.com/opencv/opencv/wiki/How_to_contribute).

 This is a template helping you to create an issue which can be processed as quickly as possible. This is the bug reporting section for the OpenCV library.
@@ -27,4 +27,4 @@ This is a template helping you to create an issue which can be processed as quic
    // C++ code example
    ```
 or attach as .txt or .zip file
-->
+-->
@@ -1778,30 +1778,30 @@ TegraCvtColor_Invoker(bgrx2hsvf, bgrx2hsv, src_data + static_cast<size_t>(range.
    : CV_HAL_ERROR_NOT_IMPLEMENTED \
 )

-#define TEGRA_CVT2PYUVTOBGR(src_data, src_step, dst_data, dst_step, dst_width, dst_height, dcn, swapBlue, uIdx) \
+#define TEGRA_CVT2PYUVTOBGR_EX(y_data, y_step, uv_data, uv_step, dst_data, dst_step, dst_width, dst_height, dcn, swapBlue, uIdx) \
 ( \
    CAROTENE_NS::isSupportedConfiguration() ? \
        dcn == 3 ? \
            uIdx == 0 ? \
                (swapBlue ? \
                    CAROTENE_NS::yuv420i2rgb(CAROTENE_NS::Size2D(dst_width, dst_height), \
-                                             src_data, src_step, \
-                                             src_data + src_step * dst_height, src_step, \
+                                             y_data, y_step, \
+                                             uv_data, uv_step, \
                                             dst_data, dst_step) : \
                    CAROTENE_NS::yuv420i2bgr(CAROTENE_NS::Size2D(dst_width, dst_height), \
-                                             src_data, src_step, \
-                                             src_data + src_step * dst_height, src_step, \
+                                             y_data, y_step, \
+                                             uv_data, uv_step, \
                                             dst_data, dst_step)), \
                CV_HAL_ERROR_OK : \
            uIdx == 1 ? \
                (swapBlue ? \
                    CAROTENE_NS::yuv420sp2rgb(CAROTENE_NS::Size2D(dst_width, dst_height), \
-                                              src_data, src_step, \
-                                              src_data + src_step * dst_height, src_step, \
+                                              y_data, y_step, \
+                                              uv_data, uv_step, \
                                              dst_data, dst_step) : \
                    CAROTENE_NS::yuv420sp2bgr(CAROTENE_NS::Size2D(dst_width, dst_height), \
-                                              src_data, src_step, \
-                                              src_data + src_step * dst_height, src_step, \
+                                              y_data, y_step, \
+                                              uv_data, uv_step, \
                                              dst_data, dst_step)), \
                CV_HAL_ERROR_OK : \
            CV_HAL_ERROR_NOT_IMPLEMENTED : \
@@ -1809,29 +1809,32 @@ TegraCvtColor_Invoker(bgrx2hsvf, bgrx2hsv, src_data + static_cast<size_t>(range.
            uIdx == 0 ? \
                (swapBlue ? \
                    CAROTENE_NS::yuv420i2rgbx(CAROTENE_NS::Size2D(dst_width, dst_height), \
-                                              src_data, src_step, \
-                                              src_data + src_step * dst_height, src_step, \
+                                              y_data, y_step, \
+                                              uv_data, uv_step, \
                                              dst_data, dst_step) : \
                    CAROTENE_NS::yuv420i2bgrx(CAROTENE_NS::Size2D(dst_width, dst_height), \
-                                              src_data, src_step, \
-                                              src_data + src_step * dst_height, src_step, \
+                                              y_data, y_step, \
+                                              uv_data, uv_step, \
                                              dst_data, dst_step)), \
                CV_HAL_ERROR_OK : \
            uIdx == 1 ? \
                (swapBlue ? \
                    CAROTENE_NS::yuv420sp2rgbx(CAROTENE_NS::Size2D(dst_width, dst_height), \
-                                               src_data, src_step, \
-                                               src_data + src_step * dst_height, src_step, \
+                                               y_data, y_step, \
+                                               uv_data, uv_step, \
                                               dst_data, dst_step) : \
                    CAROTENE_NS::yuv420sp2bgrx(CAROTENE_NS::Size2D(dst_width, dst_height), \
-                                               src_data, src_step, \
-                                               src_data + src_step * dst_height, src_step, \
+                                               y_data, y_step, \
+                                               uv_data, uv_step, \
                                               dst_data, dst_step)), \
                CV_HAL_ERROR_OK : \
            CV_HAL_ERROR_NOT_IMPLEMENTED : \
        CV_HAL_ERROR_NOT_IMPLEMENTED \
    : CV_HAL_ERROR_NOT_IMPLEMENTED \
 )
+#define TEGRA_CVT2PYUVTOBGR(src_data, src_step, dst_data, dst_step, dst_width, dst_height, dcn, swapBlue, uIdx) \
+    TEGRA_CVT2PYUVTOBGR_EX(src_data, src_step, src_data + src_step * dst_height, src_step, dst_data, dst_step, \
+            dst_width, dst_height, dcn, swapBlue, uIdx);

 #undef cv_hal_cvtBGRtoBGR
 #define cv_hal_cvtBGRtoBGR TEGRA_CVTBGRTOBGR
@@ -1841,12 +1844,18 @@ TegraCvtColor_Invoker(bgrx2hsvf, bgrx2hsv, src_data + static_cast<size_t>(range.
 #define cv_hal_cvtBGRtoGray TEGRA_CVTBGRTOGRAY
 #undef cv_hal_cvtGraytoBGR
 #define cv_hal_cvtGraytoBGR TEGRA_CVTGRAYTOBGR
+#if 0  // bit-exact tests are failed
 #undef cv_hal_cvtBGRtoYUV
 #define cv_hal_cvtBGRtoYUV TEGRA_CVTBGRTOYUV
+#endif
 #undef cv_hal_cvtBGRtoHSV
 #define cv_hal_cvtBGRtoHSV TEGRA_CVTBGRTOHSV
+#if 0  // bit-exact tests are failed
 #undef cv_hal_cvtTwoPlaneYUVtoBGR
 #define cv_hal_cvtTwoPlaneYUVtoBGR TEGRA_CVT2PYUVTOBGR
+#undef cv_hal_cvtTwoPlaneYUVtoBGREx
+#define cv_hal_cvtTwoPlaneYUVtoBGREx TEGRA_CVT2PYUVTOBGR_EX
+#endif

 #endif // OPENCV_IMGPROC_HAL_INTERFACE_H

@@ -758,7 +758,7 @@ inline void resizeAreaRounding(const Size2D &ssize, const Size2D &dsize,
    }
    else if (channels == 3)
    {
-        if ((wr == 2.0f) && (wr == 2.0f))
+        if ((wr == 2.0f) && (hr == 2.0f))
        {
 #ifndef __ANDROID__
            size_t roiw16 = dsize.width >= 15 ? (dsize.width - 15) * 3 : 0;
@@ -1,8 +1,8 @@
-# Binaries branch name: ffmpeg/3.4_20200907
-# Binaries were created for OpenCV: 03bee14372f5537daa56c62e771ec16181ca1f98
-ocv_update(FFMPEG_BINARIES_COMMIT "2a96257b743695a47f8012aab1ffb995a1dee8b4")
-ocv_update(FFMPEG_FILE_HASH_BIN32 "5e68a3ff82f43ac6524e50e448a34c9c")
-ocv_update(FFMPEG_FILE_HASH_BIN64 "205db629d893e7d4865fd1459807ff47")
+# Binaries branch name: ffmpeg/3.4_20211220
+# Binaries were created for OpenCV: a22dd28e0272ec0f1cfee8811d3f5f0392827c65
+ocv_update(FFMPEG_BINARIES_COMMIT "5a7644ec3940c6eed41c6ebb5a0602a5615fdb3f")
+ocv_update(FFMPEG_FILE_HASH_BIN32 "8ad9de6f1f2ca77786748d1f3a4e83ea")
+ocv_update(FFMPEG_FILE_HASH_BIN64 "2c670f068252e7cd28d3883993dc1d6e")
 ocv_update(FFMPEG_FILE_HASH_CMAKE "3b90f67f4b429e77d3da36698cef700c")

 function(download_win_ffmpeg script_var)
@@ -3,10 +3,10 @@ project(${JPEG_LIBRARY} C)
 ocv_warnings_disable(CMAKE_C_FLAGS -Wunused-parameter -Wsign-compare -Wshorten-64-to-32 -Wimplicit-fallthrough)

 set(VERSION_MAJOR 2)
-set(VERSION_MINOR 0)
-set(VERSION_REVISION 6)
+set(VERSION_MINOR 1)
+set(VERSION_REVISION 2)
 set(VERSION ${VERSION_MAJOR}.${VERSION_MINOR}.${VERSION_REVISION})
-set(LIBJPEG_TURBO_VERSION_NUMBER 2000006)
+set(LIBJPEG_TURBO_VERSION_NUMBER 2001002)

 string(TIMESTAMP BUILD "opencv-${OPENCV_VERSION}-libjpeg-turbo")
 if(CMAKE_BUILD_TYPE STREQUAL "Debug")
@@ -46,7 +46,6 @@ if(UNIX)
  ocv_update(HAVE_UNSIGNED_SHORT 1)
  # undef INCOMPLETE_TYPES_BROKEN
  ocv_update(RIGHT_SHIFT_IS_UNSIGNED 0)
-  ocv_update(__CHAR_UNSIGNED__ 0)
 endif()


@@ -91,7 +91,7 @@ best of our understanding.
 The Modified (3-clause) BSD License
 ===================================

-Copyright (C)2009-2020 D. R. Commander.  All Rights Reserved.
+Copyright (C)2009-2021 D. R. Commander.  All Rights Reserved.<br>
 Copyright (C)2015 Viktor Szathmáry.  All Rights Reserved.

 Redistribution and use in source and binary forms, with or without
@@ -128,7 +128,7 @@ with respect to this software, its quality, accuracy, merchantability, or
 fitness for a particular purpose.  This software is provided "AS IS", and you,
 its user, assume the entire risk as to its quality and accuracy.

-This software is copyright (C) 1991-2016, Thomas G. Lane, Guido Vollbeding.
+This software is copyright (C) 1991-2020, Thomas G. Lane, Guido Vollbeding.
 All Rights Reserved except as specified below.

 Permission is hereby granted to use, copy, modify, and distribute this
@@ -159,19 +159,6 @@ commercial products, provided that all warranty or liability claims are
 assumed by the product vendor.


-The IJG distribution formerly included code to read and write GIF files.
-To avoid entanglement with the Unisys LZW patent (now expired), GIF reading
-support has been removed altogether, and the GIF writer has been simplified
-to produce "uncompressed GIFs".  This technique does not use the LZW
-algorithm; the resulting GIF files are larger than usual, but are readable
-by all standard GIF decoders.
-
-We are required to state that
-    "The Graphics Interchange Format(c) is the Copyright property of
-    CompuServe Incorporated.  GIF(sm) is a Service Mark property of
-    CompuServe Incorporated."
-
-
 REFERENCES
 ==========

@@ -3,7 +3,7 @@ Background

 libjpeg-turbo is a JPEG image codec that uses SIMD instructions to accelerate
 baseline JPEG compression and decompression on x86, x86-64, Arm, PowerPC, and
-MIPS systems, as well as progressive JPEG compression on x86 and x86-64
+MIPS systems, as well as progressive JPEG compression on x86, x86-64, and Arm
 systems.  On such systems, libjpeg-turbo is generally 2-6x as fast as libjpeg,
 all else being equal.  On other types of systems, libjpeg-turbo can still
 outperform libjpeg by a significant amount, by virtue of its highly-optimized
@@ -61,11 +61,6 @@
   unsigned. */
 #cmakedefine RIGHT_SHIFT_IS_UNSIGNED 1

-/* Define to 1 if type `char' is unsigned and you are not using gcc.  */
-#ifndef __CHAR_UNSIGNED__
-  #cmakedefine __CHAR_UNSIGNED__ 1
-#endif
-
 /* Define to empty if `const' does not conform to ANSI C. */
 /* #undef const */

@@ -18,7 +18,6 @@
 #define HAVE_UNSIGNED_SHORT
 #undef INCOMPLETE_TYPES_BROKEN
 #undef RIGHT_SHIFT_IS_UNSIGNED
-#undef __CHAR_UNSIGNED__

 /* Define "boolean" as unsigned char, not int, per Windows custom */
 #ifndef __RPCNDR_H__            /* don't conflict if rpcndr.h already read */
@@ -40,3 +40,13 @@
 #define HAVE_BITSCANFORWARD
 #endif
 #endif
+
+#if defined(__has_attribute)
+#if __has_attribute(fallthrough)
+#define FALLTHROUGH  __attribute__((fallthrough));
+#else
+#define FALLTHROUGH
+#endif
+#else
+#define FALLTHROUGH
+#endif
@@ -48,9 +48,9 @@ rgb_ycc_convert_internal(j_compress_ptr cinfo, JSAMPARRAY input_buf,
    outptr2 = output_buf[2][output_row];
    output_row++;
    for (col = 0; col < num_cols; col++) {
-      r = GETJSAMPLE(inptr[RGB_RED]);
-      g = GETJSAMPLE(inptr[RGB_GREEN]);
-      b = GETJSAMPLE(inptr[RGB_BLUE]);
+      r = inptr[RGB_RED];
+      g = inptr[RGB_GREEN];
+      b = inptr[RGB_BLUE];
      inptr += RGB_PIXELSIZE;
      /* If the inputs are 0..MAXJSAMPLE, the outputs of these equations
       * must be too; we do not need an explicit range-limiting operation.
@@ -100,9 +100,9 @@ rgb_gray_convert_internal(j_compress_ptr cinfo, JSAMPARRAY input_buf,
    outptr = output_buf[0][output_row];
    output_row++;
    for (col = 0; col < num_cols; col++) {
-      r = GETJSAMPLE(inptr[RGB_RED]);
-      g = GETJSAMPLE(inptr[RGB_GREEN]);
-      b = GETJSAMPLE(inptr[RGB_BLUE]);
+      r = inptr[RGB_RED];
+      g = inptr[RGB_GREEN];
+      b = inptr[RGB_BLUE];
      inptr += RGB_PIXELSIZE;
      /* Y */
      outptr[col] = (JSAMPLE)((ctab[r + R_Y_OFF] + ctab[g + G_Y_OFF] +
@@ -135,9 +135,9 @@ rgb_rgb_convert_internal(j_compress_ptr cinfo, JSAMPARRAY input_buf,
    outptr2 = output_buf[2][output_row];
    output_row++;
    for (col = 0; col < num_cols; col++) {
-      outptr0[col] = GETJSAMPLE(inptr[RGB_RED]);
-      outptr1[col] = GETJSAMPLE(inptr[RGB_GREEN]);
-      outptr2[col] = GETJSAMPLE(inptr[RGB_BLUE]);
+      outptr0[col] = inptr[RGB_RED];
+      outptr1[col] = inptr[RGB_GREEN];
+      outptr2[col] = inptr[RGB_BLUE];
      inptr += RGB_PIXELSIZE;
    }
  }
@@ -392,11 +392,11 @@ cmyk_ycck_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf,
    outptr3 = output_buf[3][output_row];
    output_row++;
    for (col = 0; col < num_cols; col++) {
-      r = MAXJSAMPLE - GETJSAMPLE(inptr[0]);
-      g = MAXJSAMPLE - GETJSAMPLE(inptr[1]);
-      b = MAXJSAMPLE - GETJSAMPLE(inptr[2]);
+      r = MAXJSAMPLE - inptr[0];
+      g = MAXJSAMPLE - inptr[1];
+      b = MAXJSAMPLE - inptr[2];
      /* K passes through as-is */
-      outptr3[col] = inptr[3];  /* don't need GETJSAMPLE here */
+      outptr3[col] = inptr[3];
      inptr += 4;
      /* If the inputs are 0..MAXJSAMPLE, the outputs of these equations
       * must be too; we do not need an explicit range-limiting operation.
@@ -438,7 +438,7 @@ grayscale_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf,
    outptr = output_buf[0][output_row];
    output_row++;
    for (col = 0; col < num_cols; col++) {
-      outptr[col] = inptr[0];   /* don't need GETJSAMPLE() here */
+      outptr[col] = inptr[0];
      inptr += instride;
    }
  }
@@ -497,7 +497,7 @@ null_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
        inptr = *input_buf;
        outptr = output_buf[ci][output_row];
        for (col = 0; col < num_cols; col++) {
-          outptr[col] = inptr[ci]; /* don't need GETJSAMPLE() here */
+          outptr[col] = inptr[ci];
          inptr += nc;
        }
      }
@@ -381,19 +381,19 @@ convsamp(JSAMPARRAY sample_data, JDIMENSION start_col, DCTELEM *workspace)
    elemptr = sample_data[elemr] + start_col;

 #if DCTSIZE == 8                /* unroll the inner loop */
-    *workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
-    *workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
-    *workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
-    *workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
-    *workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
-    *workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
-    *workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
-    *workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
+    *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
+    *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
+    *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
+    *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
+    *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
+    *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
+    *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
+    *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
 #else
    {
      register int elemc;
      for (elemc = DCTSIZE; elemc > 0; elemc--)
-        *workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
+        *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
    }
 #endif
  }
@@ -533,20 +533,19 @@ convsamp_float(JSAMPARRAY sample_data, JDIMENSION start_col,
  for (elemr = 0; elemr < DCTSIZE; elemr++) {
    elemptr = sample_data[elemr] + start_col;
 #if DCTSIZE == 8                /* unroll the inner loop */
-    *workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
-    *workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
-    *workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
-    *workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
-    *workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
-    *workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
-    *workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
-    *workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
+    *workspaceptr++ = (FAST_FLOAT)((*elemptr++) - CENTERJSAMPLE);
+    *workspaceptr++ = (FAST_FLOAT)((*elemptr++) - CENTERJSAMPLE);
+    *workspaceptr++ = (FAST_FLOAT)((*elemptr++) - CENTERJSAMPLE);
+    *workspaceptr++ = (FAST_FLOAT)((*elemptr++) - CENTERJSAMPLE);
+    *workspaceptr++ = (FAST_FLOAT)((*elemptr++) - CENTERJSAMPLE);
+    *workspaceptr++ = (FAST_FLOAT)((*elemptr++) - CENTERJSAMPLE);
+    *workspaceptr++ = (FAST_FLOAT)((*elemptr++) - CENTERJSAMPLE);
+    *workspaceptr++ = (FAST_FLOAT)((*elemptr++) - CENTERJSAMPLE);
 #else
    {
      register int elemc;
      for (elemc = DCTSIZE; elemc > 0; elemc--)
-        *workspaceptr++ = (FAST_FLOAT)
-                          (GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
+        *workspaceptr++ = (FAST_FLOAT)((*elemptr++) - CENTERJSAMPLE);
    }
 #endif
  }
@@ -4,8 +4,10 @@
 * This file was part of the Independent JPEG Group's software:
 * Copyright (C) 1991-1997, Thomas G. Lane.
 * libjpeg-turbo Modifications:
- * Copyright (C) 2009-2011, 2014-2016, 2018-2019, D. R. Commander.
+ * Copyright (C) 2009-2011, 2014-2016, 2018-2021, D. R. Commander.
 * Copyright (C) 2015, Matthieu Darbois.
+ * Copyright (C) 2018, Matthias Räncker.
+ * Copyright (C) 2020, Arm Limited.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -43,14 +45,19 @@
 */

 /* NOTE: Both GCC and Clang define __GNUC__ */
-#if defined(__GNUC__) && (defined(__arm__) || defined(__aarch64__))
+#if (defined(__GNUC__) && (defined(__arm__) || defined(__aarch64__))) || \
+    defined(_M_ARM) || defined(_M_ARM64)
 #if !defined(__thumb__) || defined(__thumb2__)
 #define USE_CLZ_INTRINSIC
 #endif
 #endif

 #ifdef USE_CLZ_INTRINSIC
+#if defined(_MSC_VER) && !defined(__clang__)
+#define JPEG_NBITS_NONZERO(x)  (32 - _CountLeadingZeros(x))
+#else
 #define JPEG_NBITS_NONZERO(x)  (32 - __builtin_clz(x))
+#endif
 #define JPEG_NBITS(x)          (x ? JPEG_NBITS_NONZERO(x) : 0)
 #else
 #include "jpeg_nbits_table.h"
@@ -65,32 +72,43 @@
 * but must not be updated permanently until we complete the MCU.
 */

+#if defined(__x86_64__) && defined(__ILP32__)
+typedef unsigned long long bit_buf_type;
+#else
+typedef size_t bit_buf_type;
+#endif
+
+/* NOTE: The more optimal Huffman encoding algorithm is only used by the
+ * intrinsics implementation of the Arm Neon SIMD extensions, which is why we
+ * retain the old Huffman encoder behavior when using the GAS implementation.
+ */
+#if defined(WITH_SIMD) && !(defined(__arm__) || defined(__aarch64__) || \
+                            defined(_M_ARM) || defined(_M_ARM64))
+typedef unsigned long long simd_bit_buf_type;
+#else
+typedef bit_buf_type simd_bit_buf_type;
+#endif
+
+#if (defined(SIZEOF_SIZE_T) && SIZEOF_SIZE_T == 8) || defined(_WIN64) || \
+    (defined(__x86_64__) && defined(__ILP32__))
+#define BIT_BUF_SIZE  64
+#elif (defined(SIZEOF_SIZE_T) && SIZEOF_SIZE_T == 4) || defined(_WIN32)
+#define BIT_BUF_SIZE  32
+#else
+#error Cannot determine word size
+#endif
+#define SIMD_BIT_BUF_SIZE  (sizeof(simd_bit_buf_type) * 8)
+
 typedef struct {
-  size_t put_buffer;                    /* current bit-accumulation buffer */
-  int put_bits;                         /* # of bits now in it */
+  union {
+    bit_buf_type c;
+    simd_bit_buf_type simd;
+  } put_buffer;                         /* current bit accumulation buffer */
+  int free_bits;                        /* # of bits available in it */
+                                        /* (Neon GAS: # of bits now in it) */
  int last_dc_val[MAX_COMPS_IN_SCAN];   /* last DC coef for each component */
 } savable_state;

-/* This macro is to work around compilers with missing or broken
- * structure assignment.  You'll need to fix this code if you have
- * such a compiler and you change MAX_COMPS_IN_SCAN.
- */
-
-#ifndef NO_STRUCT_ASSIGN
-#define ASSIGN_STATE(dest, src)  ((dest) = (src))
-#else
-#if MAX_COMPS_IN_SCAN == 4
-#define ASSIGN_STATE(dest, src) \
-  ((dest).put_buffer = (src).put_buffer, \
-   (dest).put_bits = (src).put_bits, \
-   (dest).last_dc_val[0] = (src).last_dc_val[0], \
-   (dest).last_dc_val[1] = (src).last_dc_val[1], \
-   (dest).last_dc_val[2] = (src).last_dc_val[2], \
-   (dest).last_dc_val[3] = (src).last_dc_val[3])
-#endif
-#endif
-
-
 typedef struct {
  struct jpeg_entropy_encoder pub; /* public fields */

@@ -123,6 +141,7 @@ typedef struct {
  size_t free_in_buffer;        /* # of byte spaces remaining in buffer */
  savable_state cur;            /* Current bit buffer & DC state */
  j_compress_ptr cinfo;         /* dump_buffer needs access to this */
+  int simd;
 } working_state;


@@ -201,8 +220,17 @@ start_pass_huff(j_compress_ptr cinfo, boolean gather_statistics)
  }

  /* Initialize bit buffer to empty */
-  entropy->saved.put_buffer = 0;
-  entropy->saved.put_bits = 0;
+  if (entropy->simd) {
+    entropy->saved.put_buffer.simd = 0;
+#if defined(__aarch64__) && !defined(NEON_INTRINSICS)
+    entropy->saved.free_bits = 0;
+#else
+    entropy->saved.free_bits = SIMD_BIT_BUF_SIZE;
+#endif
+  } else {
+    entropy->saved.put_buffer.c = 0;
+    entropy->saved.free_bits = BIT_BUF_SIZE;
+  }

  /* Initialize restart stuff */
  entropy->restarts_to_go = cinfo->restart_interval;
@@ -287,6 +315,7 @@ jpeg_make_c_derived_tbl(j_compress_ptr cinfo, boolean isDC, int tblno,
   * this lets us detect duplicate VAL entries here, and later
   * allows emit_bits to detect any attempt to emit such symbols.
   */
+  MEMZERO(dtbl->ehufco, sizeof(dtbl->ehufco));
  MEMZERO(dtbl->ehufsi, sizeof(dtbl->ehufsi));

  /* This is also a convenient place to check for out-of-range
@@ -334,94 +363,94 @@ dump_buffer(working_state *state)

 /* Outputting bits to the file */

-/* These macros perform the same task as the emit_bits() function in the
- * original libjpeg code.  In addition to reducing overhead by explicitly
- * inlining the code, additional performance is achieved by taking into
- * account the size of the bit buffer and waiting until it is almost full
- * before emptying it.  This mostly benefits 64-bit platforms, since 6
- * bytes can be stored in a 64-bit bit buffer before it has to be emptied.
+/* Output byte b and, speculatively, an additional 0 byte.  0xFF must be
+ * encoded as 0xFF 0x00, so the output buffer pointer is advanced by 2 if the
+ * byte is 0xFF.  Otherwise, the output buffer pointer is advanced by 1, and
+ * the speculative 0 byte will be overwritten by the next byte.
 */
-
-#define EMIT_BYTE() { \
-  JOCTET c; \
-  put_bits -= 8; \
-  c = (JOCTET)GETJOCTET(put_buffer >> put_bits); \
-  *buffer++ = c; \
-  if (c == 0xFF)  /* need to stuff a zero byte? */ \
-    *buffer++ = 0; \
+#define EMIT_BYTE(b) { \
+  buffer[0] = (JOCTET)(b); \
+  buffer[1] = 0; \
+  buffer -= -2 + ((JOCTET)(b) < 0xFF); \
 }

-#define PUT_BITS(code, size) { \
-  put_bits += size; \
-  put_buffer = (put_buffer << size) | code; \
-}
+/* Output the entire bit buffer.  If there are no 0xFF bytes in it, then write
+ * directly to the output buffer.  Otherwise, use the EMIT_BYTE() macro to
+ * encode 0xFF as 0xFF 0x00.
+ */
+#if BIT_BUF_SIZE == 64

-#if SIZEOF_SIZE_T != 8 && !defined(_WIN64)
-
-#define CHECKBUF15() { \
-  if (put_bits > 15) { \
-    EMIT_BYTE() \
-    EMIT_BYTE() \
+#define FLUSH() { \
+  if (put_buffer & 0x8080808080808080 & ~(put_buffer + 0x0101010101010101)) { \
+    EMIT_BYTE(put_buffer >> 56) \
+    EMIT_BYTE(put_buffer >> 48) \
+    EMIT_BYTE(put_buffer >> 40) \
+    EMIT_BYTE(put_buffer >> 32) \
+    EMIT_BYTE(put_buffer >> 24) \
+    EMIT_BYTE(put_buffer >> 16) \
+    EMIT_BYTE(put_buffer >>  8) \
+    EMIT_BYTE(put_buffer      ) \
+  } else { \
+    buffer[0] = (JOCTET)(put_buffer >> 56); \
+    buffer[1] = (JOCTET)(put_buffer >> 48); \
+    buffer[2] = (JOCTET)(put_buffer >> 40); \
+    buffer[3] = (JOCTET)(put_buffer >> 32); \
+    buffer[4] = (JOCTET)(put_buffer >> 24); \
+    buffer[5] = (JOCTET)(put_buffer >> 16); \
+    buffer[6] = (JOCTET)(put_buffer >> 8); \
+    buffer[7] = (JOCTET)(put_buffer); \
+    buffer += 8; \
  } \
 }

-#endif
-
-#define CHECKBUF31() { \
-  if (put_bits > 31) { \
-    EMIT_BYTE() \
-    EMIT_BYTE() \
-    EMIT_BYTE() \
-    EMIT_BYTE() \
-  } \
-}
-
-#define CHECKBUF47() { \
-  if (put_bits > 47) { \
-    EMIT_BYTE() \
-    EMIT_BYTE() \
-    EMIT_BYTE() \
-    EMIT_BYTE() \
-    EMIT_BYTE() \
-    EMIT_BYTE() \
-  } \
-}
-
-#if !defined(_WIN32) && !defined(SIZEOF_SIZE_T)
-#error Cannot determine word size
-#endif
-
-#if SIZEOF_SIZE_T == 8 || defined(_WIN64)
-
-#define EMIT_BITS(code, size) { \
-  CHECKBUF47() \
-  PUT_BITS(code, size) \
-}
-
-#define EMIT_CODE(code, size) { \
-  temp2 &= (((JLONG)1) << nbits) - 1; \
-  CHECKBUF31() \
-  PUT_BITS(code, size) \
-  PUT_BITS(temp2, nbits) \
-}
-
 #else

-#define EMIT_BITS(code, size) { \
-  PUT_BITS(code, size) \
-  CHECKBUF15() \
-}
-
-#define EMIT_CODE(code, size) { \
-  temp2 &= (((JLONG)1) << nbits) - 1; \
-  PUT_BITS(code, size) \
-  CHECKBUF15() \
-  PUT_BITS(temp2, nbits) \
-  CHECKBUF15() \
+#define FLUSH() { \
+  if (put_buffer & 0x80808080 & ~(put_buffer + 0x01010101)) { \
+    EMIT_BYTE(put_buffer >> 24) \
+    EMIT_BYTE(put_buffer >> 16) \
+    EMIT_BYTE(put_buffer >>  8) \
+    EMIT_BYTE(put_buffer      ) \
+  } else { \
+    buffer[0] = (JOCTET)(put_buffer >> 24); \
+    buffer[1] = (JOCTET)(put_buffer >> 16); \
+    buffer[2] = (JOCTET)(put_buffer >> 8); \
+    buffer[3] = (JOCTET)(put_buffer); \
+    buffer += 4; \
+  } \
 }

 #endif

+/* Fill the bit buffer to capacity with the leading bits from code, then output
+ * the bit buffer and put the remaining bits from code into the bit buffer.
+ */
+#define PUT_AND_FLUSH(code, size) { \
+  put_buffer = (put_buffer << (size + free_bits)) | (code >> -free_bits); \
+  FLUSH() \
+  free_bits += BIT_BUF_SIZE; \
+  put_buffer = code; \
+}
+
+/* Insert code into the bit buffer and output the bit buffer if needed.
+ * NOTE: We can't flush with free_bits == 0, since the left shift in
+ * PUT_AND_FLUSH() would have undefined behavior.
+ */
+#define PUT_BITS(code, size) { \
+  free_bits -= size; \
+  if (free_bits < 0) \
+    PUT_AND_FLUSH(code, size) \
+  else \
+    put_buffer = (put_buffer << size) | code; \
+}
+
+#define PUT_CODE(code, size) { \
+  temp &= (((JLONG)1) << nbits) - 1; \
+  temp |= code << nbits; \
+  nbits += size; \
+  PUT_BITS(temp, nbits) \
+}
+

 /* Although it is exceedingly rare, it is possible for a Huffman-encoded
 * coefficient block to be larger than the 128-byte unencoded block.  For each
@@ -444,6 +473,7 @@ dump_buffer(working_state *state)

 #define STORE_BUFFER() { \
  if (localbuf) { \
+    size_t bytes, bytestocopy; \
    bytes = buffer - _buffer; \
    buffer = _buffer; \
    while (bytes > 0) { \
@@ -466,20 +496,46 @@ dump_buffer(working_state *state)
 LOCAL(boolean)
 flush_bits(working_state *state)
 {
-  JOCTET _buffer[BUFSIZE], *buffer;
-  size_t put_buffer;  int put_bits;
-  size_t bytes, bytestocopy;  int localbuf = 0;
+  JOCTET _buffer[BUFSIZE], *buffer, temp;
+  simd_bit_buf_type put_buffer;  int put_bits;
+  int localbuf = 0;
+
+  if (state->simd) {
+#if defined(__aarch64__) && !defined(NEON_INTRINSICS)
+    put_bits = state->cur.free_bits;
+#else
+    put_bits = SIMD_BIT_BUF_SIZE - state->cur.free_bits;
+#endif
+    put_buffer = state->cur.put_buffer.simd;
+  } else {
+    put_bits = BIT_BUF_SIZE - state->cur.free_bits;
+    put_buffer = state->cur.put_buffer.c;
+  }

-  put_buffer = state->cur.put_buffer;
-  put_bits = state->cur.put_bits;
  LOAD_BUFFER()

-  /* fill any partial byte with ones */
-  PUT_BITS(0x7F, 7)
-  while (put_bits >= 8) EMIT_BYTE()
+  while (put_bits >= 8) {
+    put_bits -= 8;
+    temp = (JOCTET)(put_buffer >> put_bits);
+    EMIT_BYTE(temp)
+  }
+  if (put_bits) {
+    /* fill partial byte with ones */
+    temp = (JOCTET)((put_buffer << (8 - put_bits)) | (0xFF >> put_bits));
+    EMIT_BYTE(temp)
+  }

-  state->cur.put_buffer = 0;    /* and reset bit-buffer to empty */
-  state->cur.put_bits = 0;
+  if (state->simd) {                    /* and reset bit buffer to empty */
+    state->cur.put_buffer.simd = 0;
+#if defined(__aarch64__) && !defined(NEON_INTRINSICS)
+    state->cur.free_bits = 0;
+#else
+    state->cur.free_bits = SIMD_BIT_BUF_SIZE;
+#endif
+  } else {
+    state->cur.put_buffer.c = 0;
+    state->cur.free_bits = BIT_BUF_SIZE;
+  }
  STORE_BUFFER()

  return TRUE;
@@ -493,7 +549,7 @@ encode_one_block_simd(working_state *state, JCOEFPTR block, int last_dc_val,
                      c_derived_tbl *dctbl, c_derived_tbl *actbl)
 {
  JOCTET _buffer[BUFSIZE], *buffer;
-  size_t bytes, bytestocopy;  int localbuf = 0;
+  int localbuf = 0;

  LOAD_BUFFER()

@@ -509,53 +565,41 @@ LOCAL(boolean)
 encode_one_block(working_state *state, JCOEFPTR block, int last_dc_val,
                 c_derived_tbl *dctbl, c_derived_tbl *actbl)
 {
-  int temp, temp2, temp3;
-  int nbits;
-  int r, code, size;
+  int temp, nbits, free_bits;
+  bit_buf_type put_buffer;
  JOCTET _buffer[BUFSIZE], *buffer;
-  size_t put_buffer;  int put_bits;
-  int code_0xf0 = actbl->ehufco[0xf0], size_0xf0 = actbl->ehufsi[0xf0];
-  size_t bytes, bytestocopy;  int localbuf = 0;
+  int localbuf = 0;

-  put_buffer = state->cur.put_buffer;
-  put_bits = state->cur.put_bits;
+  free_bits = state->cur.free_bits;
+  put_buffer = state->cur.put_buffer.c;
  LOAD_BUFFER()

  /* Encode the DC coefficient difference per section F.1.2.1 */

-  temp = temp2 = block[0] - last_dc_val;
+  temp = block[0] - last_dc_val;

  /* This is a well-known technique for obtaining the absolute value without a
   * branch.  It is derived from an assembly language technique presented in
   * "How to Optimize for the Pentium Processors", Copyright (c) 1996, 1997 by
-   * Agner Fog.
+   * Agner Fog.  This code assumes we are on a two's complement machine.
   */
-  temp3 = temp >> (CHAR_BIT * sizeof(int) - 1);
-  temp ^= temp3;
-  temp -= temp3;
-
-  /* For a negative input, want temp2 = bitwise complement of abs(input) */
-  /* This code assumes we are on a two's complement machine */
-  temp2 += temp3;
+  nbits = temp >> (CHAR_BIT * sizeof(int) - 1);
+  temp += nbits;
+  nbits ^= temp;

  /* Find the number of bits needed for the magnitude of the coefficient */
-  nbits = JPEG_NBITS(temp);
+  nbits = JPEG_NBITS(nbits);

-  /* Emit the Huffman-coded symbol for the number of bits */
-  code = dctbl->ehufco[nbits];
-  size = dctbl->ehufsi[nbits];
-  EMIT_BITS(code, size)
-
-  /* Mask off any extra bits in code */
-  temp2 &= (((JLONG)1) << nbits) - 1;
-
-  /* Emit that number of bits of the value, if positive, */
-  /* or the complement of its magnitude, if negative. */
-  EMIT_BITS(temp2, nbits)
+  /* Emit the Huffman-coded symbol for the number of bits.
+   * Emit that number of bits of the value, if positive,
+   * or the complement of its magnitude, if negative.
+   */
+  PUT_CODE(dctbl->ehufco[nbits], dctbl->ehufsi[nbits])

  /* Encode the AC coefficients per section F.1.2.2 */

-  r = 0;                        /* r = run length of zeros */
+  {
+    int r = 0;                  /* r = run length of zeros */

 /* Manually unroll the k loop to eliminate the counter variable.  This
 * improves performance greatly on systems with a limited number of
@@ -563,51 +607,46 @@ encode_one_block(working_state *state, JCOEFPTR block, int last_dc_val,
 */
 #define kloop(jpeg_natural_order_of_k) { \
  if ((temp = block[jpeg_natural_order_of_k]) == 0) { \
-    r++; \
+    r += 16; \
  } else { \
-    temp2 = temp; \
    /* Branch-less absolute value, bitwise complement, etc., same as above */ \
-    temp3 = temp >> (CHAR_BIT * sizeof(int) - 1); \
-    temp ^= temp3; \
-    temp -= temp3; \
-    temp2 += temp3; \
-    nbits = JPEG_NBITS_NONZERO(temp); \
+    nbits = temp >> (CHAR_BIT * sizeof(int) - 1); \
+    temp += nbits; \
+    nbits ^= temp; \
+    nbits = JPEG_NBITS_NONZERO(nbits); \
    /* if run length > 15, must emit special run-length-16 codes (0xF0) */ \
-    while (r > 15) { \
-      EMIT_BITS(code_0xf0, size_0xf0) \
-      r -= 16; \
+    while (r >= 16 * 16) { \
+      r -= 16 * 16; \
+      PUT_BITS(actbl->ehufco[0xf0], actbl->ehufsi[0xf0]) \
    } \
    /* Emit Huffman symbol for run length / number of bits */ \
-    temp3 = (r << 4) + nbits; \
-    code = actbl->ehufco[temp3]; \
-    size = actbl->ehufsi[temp3]; \
-    EMIT_CODE(code, size) \
+    r += nbits; \
+    PUT_CODE(actbl->ehufco[r], actbl->ehufsi[r]) \
    r = 0; \
  } \
 }

-  /* One iteration for each value in jpeg_natural_order[] */
-  kloop(1);   kloop(8);   kloop(16);  kloop(9);   kloop(2);   kloop(3);
-  kloop(10);  kloop(17);  kloop(24);  kloop(32);  kloop(25);  kloop(18);
-  kloop(11);  kloop(4);   kloop(5);   kloop(12);  kloop(19);  kloop(26);
-  kloop(33);  kloop(40);  kloop(48);  kloop(41);  kloop(34);  kloop(27);
-  kloop(20);  kloop(13);  kloop(6);   kloop(7);   kloop(14);  kloop(21);
-  kloop(28);  kloop(35);  kloop(42);  kloop(49);  kloop(56);  kloop(57);
-  kloop(50);  kloop(43);  kloop(36);  kloop(29);  kloop(22);  kloop(15);
-  kloop(23);  kloop(30);  kloop(37);  kloop(44);  kloop(51);  kloop(58);
-  kloop(59);  kloop(52);  kloop(45);  kloop(38);  kloop(31);  kloop(39);
-  kloop(46);  kloop(53);  kloop(60);  kloop(61);  kloop(54);  kloop(47);
-  kloop(55);  kloop(62);  kloop(63);
+    /* One iteration for each value in jpeg_natural_order[] */
+    kloop(1);   kloop(8);   kloop(16);  kloop(9);   kloop(2);   kloop(3);
+    kloop(10);  kloop(17);  kloop(24);  kloop(32);  kloop(25);  kloop(18);
+    kloop(11);  kloop(4);   kloop(5);   kloop(12);  kloop(19);  kloop(26);
+    kloop(33);  kloop(40);  kloop(48);  kloop(41);  kloop(34);  kloop(27);
+    kloop(20);  kloop(13);  kloop(6);   kloop(7);   kloop(14);  kloop(21);
+    kloop(28);  kloop(35);  kloop(42);  kloop(49);  kloop(56);  kloop(57);
+    kloop(50);  kloop(43);  kloop(36);  kloop(29);  kloop(22);  kloop(15);
+    kloop(23);  kloop(30);  kloop(37);  kloop(44);  kloop(51);  kloop(58);
+    kloop(59);  kloop(52);  kloop(45);  kloop(38);  kloop(31);  kloop(39);
+    kloop(46);  kloop(53);  kloop(60);  kloop(61);  kloop(54);  kloop(47);
+    kloop(55);  kloop(62);  kloop(63);

-  /* If the last coef(s) were zero, emit an end-of-block code */
-  if (r > 0) {
-    code = actbl->ehufco[0];
-    size = actbl->ehufsi[0];
-    EMIT_BITS(code, size)
+    /* If the last coef(s) were zero, emit an end-of-block code */
+    if (r > 0) {
+      PUT_BITS(actbl->ehufco[0], actbl->ehufsi[0])
+    }
  }

-  state->cur.put_buffer = put_buffer;
-  state->cur.put_bits = put_bits;
+  state->cur.put_buffer.c = put_buffer;
+  state->cur.free_bits = free_bits;
  STORE_BUFFER()

  return TRUE;
@@ -654,8 +693,9 @@ encode_mcu_huff(j_compress_ptr cinfo, JBLOCKROW *MCU_data)
  /* Load up working state */
  state.next_output_byte = cinfo->dest->next_output_byte;
  state.free_in_buffer = cinfo->dest->free_in_buffer;
-  ASSIGN_STATE(state.cur, entropy->saved);
+  state.cur = entropy->saved;
  state.cinfo = cinfo;
+  state.simd = entropy->simd;

  /* Emit restart marker if needed */
  if (cinfo->restart_interval) {
@@ -694,7 +734,7 @@ encode_mcu_huff(j_compress_ptr cinfo, JBLOCKROW *MCU_data)
  /* Completed MCU, so update state */
  cinfo->dest->next_output_byte = state.next_output_byte;
  cinfo->dest->free_in_buffer = state.free_in_buffer;
-  ASSIGN_STATE(entropy->saved, state.cur);
+  entropy->saved = state.cur;

  /* Update restart-interval state too */
  if (cinfo->restart_interval) {
@@ -723,8 +763,9 @@ finish_pass_huff(j_compress_ptr cinfo)
  /* Load up working state ... flush_bits needs it */
  state.next_output_byte = cinfo->dest->next_output_byte;
  state.free_in_buffer = cinfo->dest->free_in_buffer;
-  ASSIGN_STATE(state.cur, entropy->saved);
+  state.cur = entropy->saved;
  state.cinfo = cinfo;
+  state.simd = entropy->simd;

  /* Flush out the last data */
  if (!flush_bits(&state))
@@ -733,7 +774,7 @@ finish_pass_huff(j_compress_ptr cinfo)
  /* Update state */
  cinfo->dest->next_output_byte = state.next_output_byte;
  cinfo->dest->free_in_buffer = state.free_in_buffer;
-  ASSIGN_STATE(entropy->saved, state.cur);
+  entropy->saved = state.cur;
 }


@@ -493,7 +493,7 @@ prepare_for_pass(j_compress_ptr cinfo)
    master->pass_type = output_pass;
    master->pass_number++;
 #endif
-    /*FALLTHROUGH*/
+    FALLTHROUGH                 /*FALLTHROUGH*/
  case output_pass:
    /* Do a data-output pass. */
    /* We need not repeat per-scan setup if prior optimization pass did it. */
@@ -4,8 +4,10 @@
 * This file was part of the Independent JPEG Group's software:
 * Copyright (C) 1995-1997, Thomas G. Lane.
 * libjpeg-turbo Modifications:
- * Copyright (C) 2011, 2015, 2018, D. R. Commander.
+ * Copyright (C) 2011, 2015, 2018, 2021, D. R. Commander.
 * Copyright (C) 2016, 2018, Matthieu Darbois.
+ * Copyright (C) 2020, Arm Limited.
+ * Copyright (C) 2021, Alex Richardson.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -52,14 +54,19 @@
 */

 /* NOTE: Both GCC and Clang define __GNUC__ */
-#if defined(__GNUC__) && (defined(__arm__) || defined(__aarch64__))
+#if (defined(__GNUC__) && (defined(__arm__) || defined(__aarch64__))) || \
+    defined(_M_ARM) || defined(_M_ARM64)
 #if !defined(__thumb__) || defined(__thumb2__)
 #define USE_CLZ_INTRINSIC
 #endif
 #endif

 #ifdef USE_CLZ_INTRINSIC
+#if defined(_MSC_VER) && !defined(__clang__)
+#define JPEG_NBITS_NONZERO(x)  (32 - _CountLeadingZeros(x))
+#else
 #define JPEG_NBITS_NONZERO(x)  (32 - __builtin_clz(x))
+#endif
 #define JPEG_NBITS(x)          (x ? JPEG_NBITS_NONZERO(x) : 0)
 #else
 #include "jpeg_nbits_table.h"
@@ -169,24 +176,26 @@ INLINE
 METHODDEF(int)
 count_zeroes(size_t *x)
 {
-  int result;
 #if defined(HAVE_BUILTIN_CTZL)
+  int result;
  result = __builtin_ctzl(*x);
  *x >>= result;
 #elif defined(HAVE_BITSCANFORWARD64)
+  unsigned long result;
  _BitScanForward64(&result, *x);
  *x >>= result;
 #elif defined(HAVE_BITSCANFORWARD)
+  unsigned long result;
  _BitScanForward(&result, *x);
  *x >>= result;
 #else
-  result = 0;
+  int result = 0;
  while ((*x & 1) == 0) {
    ++result;
    *x >>= 1;
  }
 #endif
-  return result;
+  return (int)result;
 }


@@ -672,7 +681,7 @@ encode_mcu_AC_first(j_compress_ptr cinfo, JBLOCKROW *MCU_data)
      emit_restart(entropy, entropy->next_restart_num);

 #ifdef WITH_SIMD
-  cvalue = values = (JCOEF *)PAD((size_t)values_unaligned, 16);
+  cvalue = values = (JCOEF *)PAD((JUINTPTR)values_unaligned, 16);
 #else
  /* Not using SIMD, so alignment is not needed */
  cvalue = values = values_unaligned;
@@ -860,7 +869,7 @@ encode_mcu_AC_refine_prepare(const JCOEF *block,

 #define ENCODE_COEFS_AC_REFINE(label) { \
  while (zerobits) { \
-    int idx = count_zeroes(&zerobits); \
+    idx = count_zeroes(&zerobits); \
    r += idx; \
    cabsvalue += idx; \
    signbits >>= idx; \
@@ -917,7 +926,7 @@ METHODDEF(boolean)
 encode_mcu_AC_refine(j_compress_ptr cinfo, JBLOCKROW *MCU_data)
 {
  phuff_entropy_ptr entropy = (phuff_entropy_ptr)cinfo->entropy;
-  register int temp, r;
+  register int temp, r, idx;
  char *BR_buffer;
  unsigned int BR;
  int Sl = cinfo->Se - cinfo->Ss + 1;
@@ -937,7 +946,7 @@ encode_mcu_AC_refine(j_compress_ptr cinfo, JBLOCKROW *MCU_data)
      emit_restart(entropy, entropy->next_restart_num);

 #ifdef WITH_SIMD
-  cabsvalue = absvalues = (JCOEF *)PAD((size_t)absvalues_unaligned, 16);
+  cabsvalue = absvalues = (JCOEF *)PAD((JUINTPTR)absvalues_unaligned, 16);
 #else
  /* Not using SIMD, so alignment is not needed */
  cabsvalue = absvalues = absvalues_unaligned;
@@ -968,7 +977,7 @@ encode_mcu_AC_refine(j_compress_ptr cinfo, JBLOCKROW *MCU_data)

  if (zerobits) {
    int diff = ((absvalues + DCTSIZE2 / 2) - cabsvalue);
-    int idx = count_zeroes(&zerobits);
+    idx = count_zeroes(&zerobits);
    signbits >>= idx;
    idx += diff;
    r += idx;
@@ -6,7 +6,7 @@
 * libjpeg-turbo Modifications:
 * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
 * Copyright (C) 2014, MIPS Technologies, Inc., California.
- * Copyright (C) 2015, D. R. Commander.
+ * Copyright (C) 2015, 2019, D. R. Commander.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -103,7 +103,7 @@ expand_right_edge(JSAMPARRAY image_data, int num_rows, JDIMENSION input_cols,
  if (numcols > 0) {
    for (row = 0; row < num_rows; row++) {
      ptr = image_data[row] + input_cols;
-      pixval = ptr[-1];         /* don't need GETJSAMPLE() here */
+      pixval = ptr[-1];
      for (count = numcols; count > 0; count--)
        *ptr++ = pixval;
    }
@@ -174,7 +174,7 @@ int_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
      for (v = 0; v < v_expand; v++) {
        inptr = input_data[inrow + v] + outcol_h;
        for (h = 0; h < h_expand; h++) {
-          outvalue += (JLONG)GETJSAMPLE(*inptr++);
+          outvalue += (JLONG)(*inptr++);
        }
      }
      *outptr++ = (JSAMPLE)((outvalue + numpix2) / numpix);
@@ -237,8 +237,7 @@ h2v1_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
    inptr = input_data[outrow];
    bias = 0;                   /* bias = 0,1,0,1,... for successive samples */
    for (outcol = 0; outcol < output_cols; outcol++) {
-      *outptr++ =
-        (JSAMPLE)((GETJSAMPLE(*inptr) + GETJSAMPLE(inptr[1]) + bias) >> 1);
+      *outptr++ = (JSAMPLE)((inptr[0] + inptr[1] + bias) >> 1);
      bias ^= 1;                /* 0=>1, 1=>0 */
      inptr += 2;
    }
@@ -277,8 +276,7 @@ h2v2_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
    bias = 1;                   /* bias = 1,2,1,2,... for successive samples */
    for (outcol = 0; outcol < output_cols; outcol++) {
      *outptr++ =
-        (JSAMPLE)((GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[1]) +
-                   GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1]) + bias) >> 2);
+        (JSAMPLE)((inptr0[0] + inptr0[1] + inptr1[0] + inptr1[1] + bias) >> 2);
      bias ^= 3;                /* 1=>2, 2=>1 */
      inptr0 += 2;  inptr1 += 2;
    }
@@ -337,33 +335,25 @@ h2v2_smooth_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
    below_ptr = input_data[inrow + 2];

    /* Special case for first column: pretend column -1 is same as column 0 */
-    membersum = GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[1]) +
-                GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1]);
-    neighsum = GETJSAMPLE(*above_ptr) + GETJSAMPLE(above_ptr[1]) +
-               GETJSAMPLE(*below_ptr) + GETJSAMPLE(below_ptr[1]) +
-               GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[2]) +
-               GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[2]);
+    membersum = inptr0[0] + inptr0[1] + inptr1[0] + inptr1[1];
+    neighsum = above_ptr[0] + above_ptr[1] + below_ptr[0] + below_ptr[1] +
+               inptr0[0] + inptr0[2] + inptr1[0] + inptr1[2];
    neighsum += neighsum;
-    neighsum += GETJSAMPLE(*above_ptr) + GETJSAMPLE(above_ptr[2]) +
-                GETJSAMPLE(*below_ptr) + GETJSAMPLE(below_ptr[2]);
+    neighsum += above_ptr[0] + above_ptr[2] + below_ptr[0] + below_ptr[2];
    membersum = membersum * memberscale + neighsum * neighscale;
    *outptr++ = (JSAMPLE)((membersum + 32768) >> 16);
    inptr0 += 2;  inptr1 += 2;  above_ptr += 2;  below_ptr += 2;

    for (colctr = output_cols - 2; colctr > 0; colctr--) {
      /* sum of pixels directly mapped to this output element */
-      membersum = GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[1]) +
-                  GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1]);
+      membersum = inptr0[0] + inptr0[1] + inptr1[0] + inptr1[1];
      /* sum of edge-neighbor pixels */
-      neighsum = GETJSAMPLE(*above_ptr) + GETJSAMPLE(above_ptr[1]) +
-                 GETJSAMPLE(*below_ptr) + GETJSAMPLE(below_ptr[1]) +
-                 GETJSAMPLE(inptr0[-1]) + GETJSAMPLE(inptr0[2]) +
-                 GETJSAMPLE(inptr1[-1]) + GETJSAMPLE(inptr1[2]);
+      neighsum = above_ptr[0] + above_ptr[1] + below_ptr[0] + below_ptr[1] +
+                 inptr0[-1] + inptr0[2] + inptr1[-1] + inptr1[2];
      /* The edge-neighbors count twice as much as corner-neighbors */
      neighsum += neighsum;
      /* Add in the corner-neighbors */
-      neighsum += GETJSAMPLE(above_ptr[-1]) + GETJSAMPLE(above_ptr[2]) +
-                  GETJSAMPLE(below_ptr[-1]) + GETJSAMPLE(below_ptr[2]);
+      neighsum += above_ptr[-1] + above_ptr[2] + below_ptr[-1] + below_ptr[2];
      /* form final output scaled up by 2^16 */
      membersum = membersum * memberscale + neighsum * neighscale;
      /* round, descale and output it */
@@ -372,15 +362,11 @@ h2v2_smooth_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
    }

    /* Special case for last column */
-    membersum = GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[1]) +
-                GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1]);
-    neighsum = GETJSAMPLE(*above_ptr) + GETJSAMPLE(above_ptr[1]) +
-               GETJSAMPLE(*below_ptr) + GETJSAMPLE(below_ptr[1]) +
-               GETJSAMPLE(inptr0[-1]) + GETJSAMPLE(inptr0[1]) +
-               GETJSAMPLE(inptr1[-1]) + GETJSAMPLE(inptr1[1]);
+    membersum = inptr0[0] + inptr0[1] + inptr1[0] + inptr1[1];
+    neighsum = above_ptr[0] + above_ptr[1] + below_ptr[0] + below_ptr[1] +
+               inptr0[-1] + inptr0[1] + inptr1[-1] + inptr1[1];
    neighsum += neighsum;
-    neighsum += GETJSAMPLE(above_ptr[-1]) + GETJSAMPLE(above_ptr[1]) +
-                GETJSAMPLE(below_ptr[-1]) + GETJSAMPLE(below_ptr[1]);
+    neighsum += above_ptr[-1] + above_ptr[1] + below_ptr[-1] + below_ptr[1];
    membersum = membersum * memberscale + neighsum * neighscale;
    *outptr = (JSAMPLE)((membersum + 32768) >> 16);

@@ -429,21 +415,18 @@ fullsize_smooth_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
    below_ptr = input_data[outrow + 1];

    /* Special case for first column */
-    colsum = GETJSAMPLE(*above_ptr++) + GETJSAMPLE(*below_ptr++) +
-             GETJSAMPLE(*inptr);
-    membersum = GETJSAMPLE(*inptr++);
-    nextcolsum = GETJSAMPLE(*above_ptr) + GETJSAMPLE(*below_ptr) +
-                 GETJSAMPLE(*inptr);
+    colsum = (*above_ptr++) + (*below_ptr++) + inptr[0];
+    membersum = *inptr++;
+    nextcolsum = above_ptr[0] + below_ptr[0] + inptr[0];
    neighsum = colsum + (colsum - membersum) + nextcolsum;
    membersum = membersum * memberscale + neighsum * neighscale;
    *outptr++ = (JSAMPLE)((membersum + 32768) >> 16);
    lastcolsum = colsum;  colsum = nextcolsum;

    for (colctr = output_cols - 2; colctr > 0; colctr--) {
-      membersum = GETJSAMPLE(*inptr++);
+      membersum = *inptr++;
      above_ptr++;  below_ptr++;
-      nextcolsum = GETJSAMPLE(*above_ptr) + GETJSAMPLE(*below_ptr) +
-                   GETJSAMPLE(*inptr);
+      nextcolsum = above_ptr[0] + below_ptr[0] + inptr[0];
      neighsum = lastcolsum + (colsum - membersum) + nextcolsum;
      membersum = membersum * memberscale + neighsum * neighscale;
      *outptr++ = (JSAMPLE)((membersum + 32768) >> 16);
@@ -451,7 +434,7 @@ fullsize_smooth_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
    }

    /* Special case for last column */
-    membersum = GETJSAMPLE(*inptr);
+    membersum = *inptr;
    neighsum = lastcolsum + (colsum - membersum) + colsum;
    membersum = membersum * memberscale + neighsum * neighscale;
    *outptr = (JSAMPLE)((membersum + 32768) >> 16);
@@ -23,6 +23,7 @@
 #include "jinclude.h"
 #include "jpeglib.h"
 #include "jdmaster.h"
+#include "jconfigint.h"


 /*
@@ -308,7 +309,7 @@ jpeg_consume_input(j_decompress_ptr cinfo)
    /* Initialize application's data source module */
    (*cinfo->src->init_source) (cinfo);
    cinfo->global_state = DSTATE_INHEADER;
-    /*FALLTHROUGH*/
+    FALLTHROUGH                 /*FALLTHROUGH*/
  case DSTATE_INHEADER:
    retcode = (*cinfo->inputctl->consume_input) (cinfo);
    if (retcode == JPEG_REACHED_SOS) { /* Found SOS, prepare to decompress */
@@ -4,7 +4,7 @@
 * This file was part of the Independent JPEG Group's software:
 * Copyright (C) 1994-1996, Thomas G. Lane.
 * libjpeg-turbo Modifications:
- * Copyright (C) 2010, 2015-2018, 2020, D. R. Commander.
+ * Copyright (C) 2010, 2015-2020, D. R. Commander.
 * Copyright (C) 2015, Google, Inc.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
@@ -319,6 +319,8 @@ read_and_discard_scanlines(j_decompress_ptr cinfo, JDIMENSION num_lines)
 {
  JDIMENSION n;
  my_master_ptr master = (my_master_ptr)cinfo->master;
+  JSAMPLE dummy_sample[1] = { 0 };
+  JSAMPROW dummy_row = dummy_sample;
  JSAMPARRAY scanlines = NULL;
  void (*color_convert) (j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
                         JDIMENSION input_row, JSAMPARRAY output_buf,
@@ -329,6 +331,10 @@ read_and_discard_scanlines(j_decompress_ptr cinfo, JDIMENSION num_lines)
  if (cinfo->cconvert && cinfo->cconvert->color_convert) {
    color_convert = cinfo->cconvert->color_convert;
    cinfo->cconvert->color_convert = noop_convert;
+    /* This just prevents UBSan from complaining about adding 0 to a NULL
+     * pointer.  The pointer isn't actually used.
+     */
+    scanlines = &dummy_row;
  }

  if (cinfo->cquantize && cinfo->cquantize->color_quantize) {
@@ -532,6 +538,8 @@ jpeg_skip_scanlines(j_decompress_ptr cinfo, JDIMENSION num_lines)
         * decoded coefficients.  This is ~5% faster for large subsets, but
         * it's tough to tell a difference for smaller images.
         */
+        if (!cinfo->entropy->insufficient_data)
+          cinfo->master->last_good_iMCU_row = cinfo->input_iMCU_row;
        (*cinfo->entropy->decode_mcu) (cinfo, NULL);
      }
    }
@@ -4,7 +4,7 @@
 * This file was part of the Independent JPEG Group's software:
 * Developed 1997-2015 by Guido Vollbeding.
 * libjpeg-turbo Modifications:
- * Copyright (C) 2015-2018, D. R. Commander.
+ * Copyright (C) 2015-2020, D. R. Commander.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -80,7 +80,7 @@ get_byte(j_decompress_ptr cinfo)
    if (!(*src->fill_input_buffer) (cinfo))
      ERREXIT(cinfo, JERR_CANT_SUSPEND);
  src->bytes_in_buffer--;
-  return GETJOCTET(*src->next_input_byte++);
+  return *src->next_input_byte++;
 }


@@ -665,8 +665,16 @@ bad:
    for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
      int coefi, cindex = cinfo->cur_comp_info[ci]->component_index;
      int *coef_bit_ptr = &cinfo->coef_bits[cindex][0];
+      int *prev_coef_bit_ptr =
+        &cinfo->coef_bits[cindex + cinfo->num_components][0];
      if (cinfo->Ss && coef_bit_ptr[0] < 0) /* AC without prior DC scan */
        WARNMS2(cinfo, JWRN_BOGUS_PROGRESSION, cindex, 0);
+      for (coefi = MIN(cinfo->Ss, 1); coefi <= MAX(cinfo->Se, 9); coefi++) {
+        if (cinfo->input_scan_number > 1)
+          prev_coef_bit_ptr[coefi] = coef_bit_ptr[coefi];
+        else
+          prev_coef_bit_ptr[coefi] = 0;
+      }
      for (coefi = cinfo->Ss; coefi <= cinfo->Se; coefi++) {
        int expected = (coef_bit_ptr[coefi] < 0) ? 0 : coef_bit_ptr[coefi];
        if (cinfo->Ah != expected)
@@ -727,6 +735,7 @@ bad:
  entropy->c = 0;
  entropy->a = 0;
  entropy->ct = -16;    /* force reading 2 initial bytes to fill C */
+  entropy->pub.insufficient_data = FALSE;

  /* Initialize restart counter */
  entropy->restarts_to_go = cinfo->restart_interval;
@@ -763,7 +772,7 @@ jinit_arith_decoder(j_decompress_ptr cinfo)
    int *coef_bit_ptr, ci;
    cinfo->coef_bits = (int (*)[DCTSIZE2])
      (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
-                                  cinfo->num_components * DCTSIZE2 *
+                                  cinfo->num_components * 2 * DCTSIZE2 *
                                  sizeof(int));
    coef_bit_ptr = &cinfo->coef_bits[0][0];
    for (ci = 0; ci < cinfo->num_components; ci++)
@@ -5,7 +5,7 @@
 * Copyright (C) 1994-1997, Thomas G. Lane.
 * libjpeg-turbo Modifications:
 * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * Copyright (C) 2010, 2015-2016, D. R. Commander.
+ * Copyright (C) 2010, 2015-2016, 2019-2020, D. R. Commander.
 * Copyright (C) 2015, 2020, Google, Inc.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
@@ -102,6 +102,8 @@ decompress_onepass(j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
      /* Try to fetch an MCU.  Entropy decoder expects buffer to be zeroed. */
      jzero_far((void *)coef->MCU_buffer[0],
                (size_t)(cinfo->blocks_in_MCU * sizeof(JBLOCK)));
+      if (!cinfo->entropy->insufficient_data)
+        cinfo->master->last_good_iMCU_row = cinfo->input_iMCU_row;
      if (!(*cinfo->entropy->decode_mcu) (cinfo, coef->MCU_buffer)) {
        /* Suspension forced; update state counters and exit */
        coef->MCU_vert_offset = yoffset;
@@ -227,6 +229,8 @@ consume_data(j_decompress_ptr cinfo)
          }
        }
      }
+      if (!cinfo->entropy->insufficient_data)
+        cinfo->master->last_good_iMCU_row = cinfo->input_iMCU_row;
      /* Try to fetch the MCU. */
      if (!(*cinfo->entropy->decode_mcu) (cinfo, coef->MCU_buffer)) {
        /* Suspension forced; update state counters and exit */
@@ -326,19 +330,22 @@ decompress_data(j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
 #ifdef BLOCK_SMOOTHING_SUPPORTED

 /*
- * This code applies interblock smoothing as described by section K.8
- * of the JPEG standard: the first 5 AC coefficients are estimated from
- * the DC values of a DCT block and its 8 neighboring blocks.
+ * This code applies interblock smoothing; the first 9 AC coefficients are
+ * estimated from the DC values of a DCT block and its 24 neighboring blocks.
 * We apply smoothing only for progressive JPEG decoding, and only if
 * the coefficients it can estimate are not yet known to full precision.
 */

-/* Natural-order array positions of the first 5 zigzag-order coefficients */
+/* Natural-order array positions of the first 9 zigzag-order coefficients */
 #define Q01_POS  1
 #define Q10_POS  8
 #define Q20_POS  16
 #define Q11_POS  9
 #define Q02_POS  2
+#define Q03_POS  3
+#define Q12_POS  10
+#define Q21_POS  17
+#define Q30_POS  24

 /*
 * Determine whether block smoothing is applicable and safe.
@@ -356,8 +363,8 @@ smoothing_ok(j_decompress_ptr cinfo)
  int ci, coefi;
  jpeg_component_info *compptr;
  JQUANT_TBL *qtable;
-  int *coef_bits;
-  int *coef_bits_latch;
+  int *coef_bits, *prev_coef_bits;
+  int *coef_bits_latch, *prev_coef_bits_latch;

  if (!cinfo->progressive_mode || cinfo->coef_bits == NULL)
    return FALSE;
@@ -366,34 +373,47 @@ smoothing_ok(j_decompress_ptr cinfo)
  if (coef->coef_bits_latch == NULL)
    coef->coef_bits_latch = (int *)
      (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
-                                  cinfo->num_components *
+                                  cinfo->num_components * 2 *
                                  (SAVED_COEFS * sizeof(int)));
  coef_bits_latch = coef->coef_bits_latch;
+  prev_coef_bits_latch =
+    &coef->coef_bits_latch[cinfo->num_components * SAVED_COEFS];

  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
       ci++, compptr++) {
    /* All components' quantization values must already be latched. */
    if ((qtable = compptr->quant_table) == NULL)
      return FALSE;
-    /* Verify DC & first 5 AC quantizers are nonzero to avoid zero-divide. */
+    /* Verify DC & first 9 AC quantizers are nonzero to avoid zero-divide. */
    if (qtable->quantval[0] == 0 ||
        qtable->quantval[Q01_POS] == 0 ||
        qtable->quantval[Q10_POS] == 0 ||
        qtable->quantval[Q20_POS] == 0 ||
        qtable->quantval[Q11_POS] == 0 ||
-        qtable->quantval[Q02_POS] == 0)
+        qtable->quantval[Q02_POS] == 0 ||
+        qtable->quantval[Q03_POS] == 0 ||
+        qtable->quantval[Q12_POS] == 0 ||
+        qtable->quantval[Q21_POS] == 0 ||
+        qtable->quantval[Q30_POS] == 0)
      return FALSE;
    /* DC values must be at least partly known for all components. */
    coef_bits = cinfo->coef_bits[ci];
+    prev_coef_bits = cinfo->coef_bits[ci + cinfo->num_components];
    if (coef_bits[0] < 0)
      return FALSE;
+    coef_bits_latch[0] = coef_bits[0];
    /* Block smoothing is helpful if some AC coefficients remain inaccurate. */
-    for (coefi = 1; coefi <= 5; coefi++) {
+    for (coefi = 1; coefi < SAVED_COEFS; coefi++) {
+      if (cinfo->input_scan_number > 1)
+        prev_coef_bits_latch[coefi] = prev_coef_bits[coefi];
+      else
+        prev_coef_bits_latch[coefi] = -1;
      coef_bits_latch[coefi] = coef_bits[coefi];
      if (coef_bits[coefi] != 0)
        smoothing_useful = TRUE;
    }
    coef_bits_latch += SAVED_COEFS;
+    prev_coef_bits_latch += SAVED_COEFS;
  }

  return smoothing_useful;
@@ -412,17 +432,20 @@ decompress_smooth_data(j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
  JDIMENSION block_num, last_block_column;
  int ci, block_row, block_rows, access_rows;
  JBLOCKARRAY buffer;
-  JBLOCKROW buffer_ptr, prev_block_row, next_block_row;
+  JBLOCKROW buffer_ptr, prev_prev_block_row, prev_block_row;
+  JBLOCKROW next_block_row, next_next_block_row;
  JSAMPARRAY output_ptr;
  JDIMENSION output_col;
  jpeg_component_info *compptr;
  inverse_DCT_method_ptr inverse_DCT;
-  boolean first_row, last_row;
+  boolean change_dc;
  JCOEF *workspace;
  int *coef_bits;
  JQUANT_TBL *quanttbl;
-  JLONG Q00, Q01, Q02, Q10, Q11, Q20, num;
-  int DC1, DC2, DC3, DC4, DC5, DC6, DC7, DC8, DC9;
+  JLONG Q00, Q01, Q02, Q03 = 0, Q10, Q11, Q12 = 0, Q20, Q21 = 0, Q30 = 0, num;
+  int DC01, DC02, DC03, DC04, DC05, DC06, DC07, DC08, DC09, DC10, DC11, DC12,
+      DC13, DC14, DC15, DC16, DC17, DC18, DC19, DC20, DC21, DC22, DC23, DC24,
+      DC25;
  int Al, pred;

  /* Keep a local variable to avoid looking it up more than once */
@@ -434,10 +457,10 @@ decompress_smooth_data(j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
    if (cinfo->input_scan_number == cinfo->output_scan_number) {
      /* If input is working on current scan, we ordinarily want it to
       * have completed the current row.  But if input scan is DC,
-       * we want it to keep one row ahead so that next block row's DC
+       * we want it to keep two rows ahead so that next two block rows' DC
       * values are up to date.
       */
-      JDIMENSION delta = (cinfo->Ss == 0) ? 1 : 0;
+      JDIMENSION delta = (cinfo->Ss == 0) ? 2 : 0;
      if (cinfo->input_iMCU_row > cinfo->output_iMCU_row + delta)
        break;
    }
@@ -452,34 +475,53 @@ decompress_smooth_data(j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
    if (!compptr->component_needed)
      continue;
    /* Count non-dummy DCT block rows in this iMCU row. */
-    if (cinfo->output_iMCU_row < last_iMCU_row) {
+    if (cinfo->output_iMCU_row < last_iMCU_row - 1) {
+      block_rows = compptr->v_samp_factor;
+      access_rows = block_rows * 3; /* this and next two iMCU rows */
+    } else if (cinfo->output_iMCU_row < last_iMCU_row) {
      block_rows = compptr->v_samp_factor;
      access_rows = block_rows * 2; /* this and next iMCU row */
-      last_row = FALSE;
    } else {
      /* NB: can't use last_row_height here; it is input-side-dependent! */
      block_rows = (int)(compptr->height_in_blocks % compptr->v_samp_factor);
      if (block_rows == 0) block_rows = compptr->v_samp_factor;
      access_rows = block_rows; /* this iMCU row only */
-      last_row = TRUE;
    }
    /* Align the virtual buffer for this component. */
-    if (cinfo->output_iMCU_row > 0) {
-      access_rows += compptr->v_samp_factor; /* prior iMCU row too */
+    if (cinfo->output_iMCU_row > 1) {
+      access_rows += 2 * compptr->v_samp_factor; /* prior two iMCU rows too */
+      buffer = (*cinfo->mem->access_virt_barray)
+        ((j_common_ptr)cinfo, coef->whole_image[ci],
+         (cinfo->output_iMCU_row - 2) * compptr->v_samp_factor,
+         (JDIMENSION)access_rows, FALSE);
+      buffer += 2 * compptr->v_samp_factor; /* point to current iMCU row */
+    } else if (cinfo->output_iMCU_row > 0) {
      buffer = (*cinfo->mem->access_virt_barray)
        ((j_common_ptr)cinfo, coef->whole_image[ci],
         (cinfo->output_iMCU_row - 1) * compptr->v_samp_factor,
         (JDIMENSION)access_rows, FALSE);
      buffer += compptr->v_samp_factor; /* point to current iMCU row */
-      first_row = FALSE;
    } else {
      buffer = (*cinfo->mem->access_virt_barray)
        ((j_common_ptr)cinfo, coef->whole_image[ci],
         (JDIMENSION)0, (JDIMENSION)access_rows, FALSE);
-      first_row = TRUE;
    }
-    /* Fetch component-dependent info */
-    coef_bits = coef->coef_bits_latch + (ci * SAVED_COEFS);
+    /* Fetch component-dependent info.
+     * If the current scan is incomplete, then we use the component-dependent
+     * info from the previous scan.
+     */
+    if (cinfo->output_iMCU_row > cinfo->master->last_good_iMCU_row)
+      coef_bits =
+        coef->coef_bits_latch + ((ci + cinfo->num_components) * SAVED_COEFS);
+    else
+      coef_bits = coef->coef_bits_latch + (ci * SAVED_COEFS);
+
+    /* We only do DC interpolation if no AC coefficient data is available. */
+    change_dc =
+      coef_bits[1] == -1 && coef_bits[2] == -1 && coef_bits[3] == -1 &&
+      coef_bits[4] == -1 && coef_bits[5] == -1 && coef_bits[6] == -1 &&
+      coef_bits[7] == -1 && coef_bits[8] == -1 && coef_bits[9] == -1;
+
    quanttbl = compptr->quant_table;
    Q00 = quanttbl->quantval[0];
    Q01 = quanttbl->quantval[Q01_POS];
@@ -487,27 +529,51 @@ decompress_smooth_data(j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
    Q20 = quanttbl->quantval[Q20_POS];
    Q11 = quanttbl->quantval[Q11_POS];
    Q02 = quanttbl->quantval[Q02_POS];
+    if (change_dc) {
+      Q03 = quanttbl->quantval[Q03_POS];
+      Q12 = quanttbl->quantval[Q12_POS];
+      Q21 = quanttbl->quantval[Q21_POS];
+      Q30 = quanttbl->quantval[Q30_POS];
+    }
    inverse_DCT = cinfo->idct->inverse_DCT[ci];
    output_ptr = output_buf[ci];
    /* Loop over all DCT blocks to be processed. */
    for (block_row = 0; block_row < block_rows; block_row++) {
      buffer_ptr = buffer[block_row] + cinfo->master->first_MCU_col[ci];
-      if (first_row && block_row == 0)
+
+      if (block_row > 0 || cinfo->output_iMCU_row > 0)
+        prev_block_row =
+          buffer[block_row - 1] + cinfo->master->first_MCU_col[ci];
+      else
        prev_block_row = buffer_ptr;
+
+      if (block_row > 1 || cinfo->output_iMCU_row > 1)
+        prev_prev_block_row =
+          buffer[block_row - 2] + cinfo->master->first_MCU_col[ci];
+      else
+        prev_prev_block_row = prev_block_row;
+
+      if (block_row < block_rows - 1 || cinfo->output_iMCU_row < last_iMCU_row)
+        next_block_row =
+          buffer[block_row + 1] + cinfo->master->first_MCU_col[ci];
      else
-        prev_block_row = buffer[block_row - 1] +
-                         cinfo->master->first_MCU_col[ci];
-      if (last_row && block_row == block_rows - 1)
        next_block_row = buffer_ptr;
+
+      if (block_row < block_rows - 2 ||
+          cinfo->output_iMCU_row < last_iMCU_row - 1)
+        next_next_block_row =
+          buffer[block_row + 2] + cinfo->master->first_MCU_col[ci];
      else
-        next_block_row = buffer[block_row + 1] +
-                         cinfo->master->first_MCU_col[ci];
+        next_next_block_row = next_block_row;
+
      /* We fetch the surrounding DC values using a sliding-register approach.
-       * Initialize all nine here so as to do the right thing on narrow pics.
+       * Initialize all 25 here so as to do the right thing on narrow pics.
       */
-      DC1 = DC2 = DC3 = (int)prev_block_row[0][0];
-      DC4 = DC5 = DC6 = (int)buffer_ptr[0][0];
-      DC7 = DC8 = DC9 = (int)next_block_row[0][0];
+      DC01 = DC02 = DC03 = DC04 = DC05 = (int)prev_prev_block_row[0][0];
+      DC06 = DC07 = DC08 = DC09 = DC10 = (int)prev_block_row[0][0];
+      DC11 = DC12 = DC13 = DC14 = DC15 = (int)buffer_ptr[0][0];
+      DC16 = DC17 = DC18 = DC19 = DC20 = (int)next_block_row[0][0];
+      DC21 = DC22 = DC23 = DC24 = DC25 = (int)next_next_block_row[0][0];
      output_col = 0;
      last_block_column = compptr->width_in_blocks - 1;
      for (block_num = cinfo->master->first_MCU_col[ci];
@@ -515,18 +581,39 @@ decompress_smooth_data(j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
        /* Fetch current DCT block into workspace so we can modify it. */
        jcopy_block_row(buffer_ptr, (JBLOCKROW)workspace, (JDIMENSION)1);
        /* Update DC values */
-        if (block_num < last_block_column) {
-          DC3 = (int)prev_block_row[1][0];
-          DC6 = (int)buffer_ptr[1][0];
-          DC9 = (int)next_block_row[1][0];
+        if (block_num == cinfo->master->first_MCU_col[ci] &&
+            block_num < last_block_column) {
+          DC04 = (int)prev_prev_block_row[1][0];
+          DC09 = (int)prev_block_row[1][0];
+          DC14 = (int)buffer_ptr[1][0];
+          DC19 = (int)next_block_row[1][0];
+          DC24 = (int)next_next_block_row[1][0];
        }
-        /* Compute coefficient estimates per K.8.
-         * An estimate is applied only if coefficient is still zero,
-         * and is not known to be fully accurate.
+        if (block_num + 1 < last_block_column) {
+          DC05 = (int)prev_prev_block_row[2][0];
+          DC10 = (int)prev_block_row[2][0];
+          DC15 = (int)buffer_ptr[2][0];
+          DC20 = (int)next_block_row[2][0];
+          DC25 = (int)next_next_block_row[2][0];
+        }
+        /* If DC interpolation is enabled, compute coefficient estimates using
+         * a Gaussian-like kernel, keeping the averages of the DC values.
+         *
+         * If DC interpolation is disabled, compute coefficient estimates using
+         * an algorithm similar to the one described in Section K.8 of the JPEG
+         * standard, except applied to a 5x5 window rather than a 3x3 window.
+         *
+         * An estimate is applied only if the coefficient is still zero and is
+         * not known to be fully accurate.
         */
        /* AC01 */
        if ((Al = coef_bits[1]) != 0 && workspace[1] == 0) {
-          num = 36 * Q00 * (DC4 - DC6);
+          num = Q00 * (change_dc ?
+                (-DC01 - DC02 + DC04 + DC05 - 3 * DC06 + 13 * DC07 -
+                 13 * DC09 + 3 * DC10 - 3 * DC11 + 38 * DC12 - 38 * DC14 +
+                 3 * DC15 - 3 * DC16 + 13 * DC17 - 13 * DC19 + 3 * DC20 -
+                 DC21 - DC22 + DC24 + DC25) :
+                (-7 * DC11 + 50 * DC12 - 50 * DC14 + 7 * DC15));
          if (num >= 0) {
            pred = (int)(((Q01 << 7) + num) / (Q01 << 8));
            if (Al > 0 && pred >= (1 << Al))
@@ -541,7 +628,12 @@ decompress_smooth_data(j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
        }
        /* AC10 */
        if ((Al = coef_bits[2]) != 0 && workspace[8] == 0) {
-          num = 36 * Q00 * (DC2 - DC8);
+          num = Q00 * (change_dc ?
+                (-DC01 - 3 * DC02 - 3 * DC03 - 3 * DC04 - DC05 - DC06 +
+                 13 * DC07 + 38 * DC08 + 13 * DC09 - DC10 + DC16 -
+                 13 * DC17 - 38 * DC18 - 13 * DC19 + DC20 + DC21 +
+                 3 * DC22 + 3 * DC23 + 3 * DC24 + DC25) :
+                (-7 * DC03 + 50 * DC08 - 50 * DC18 + 7 * DC23));
          if (num >= 0) {
            pred = (int)(((Q10 << 7) + num) / (Q10 << 8));
            if (Al > 0 && pred >= (1 << Al))
@@ -556,7 +648,10 @@ decompress_smooth_data(j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
        }
        /* AC20 */
        if ((Al = coef_bits[3]) != 0 && workspace[16] == 0) {
-          num = 9 * Q00 * (DC2 + DC8 - 2 * DC5);
+          num = Q00 * (change_dc ?
+                (DC03 + 2 * DC07 + 7 * DC08 + 2 * DC09 - 5 * DC12 - 14 * DC13 -
+                 5 * DC14 + 2 * DC17 + 7 * DC18 + 2 * DC19 + DC23) :
+                (-DC03 + 13 * DC08 - 24 * DC13 + 13 * DC18 - DC23));
          if (num >= 0) {
            pred = (int)(((Q20 << 7) + num) / (Q20 << 8));
            if (Al > 0 && pred >= (1 << Al))
@@ -571,7 +666,11 @@ decompress_smooth_data(j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
        }
        /* AC11 */
        if ((Al = coef_bits[4]) != 0 && workspace[9] == 0) {
-          num = 5 * Q00 * (DC1 - DC3 - DC7 + DC9);
+          num = Q00 * (change_dc ?
+                (-DC01 + DC05 + 9 * DC07 - 9 * DC09 - 9 * DC17 +
+                 9 * DC19 + DC21 - DC25) :
+                (DC10 + DC16 - 10 * DC17 + 10 * DC19 - DC02 - DC20 + DC22 -
+                 DC24 + DC04 - DC06 + 10 * DC07 - 10 * DC09));
          if (num >= 0) {
            pred = (int)(((Q11 << 7) + num) / (Q11 << 8));
            if (Al > 0 && pred >= (1 << Al))
@@ -586,7 +685,10 @@ decompress_smooth_data(j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
        }
        /* AC02 */
        if ((Al = coef_bits[5]) != 0 && workspace[2] == 0) {
-          num = 9 * Q00 * (DC4 + DC6 - 2 * DC5);
+          num = Q00 * (change_dc ?
+                (2 * DC07 - 5 * DC08 + 2 * DC09 + DC11 + 7 * DC12 - 14 * DC13 +
+                 7 * DC14 + DC15 + 2 * DC17 - 5 * DC18 + 2 * DC19) :
+                (-DC11 + 13 * DC12 - 24 * DC13 + 13 * DC14 - DC15));
          if (num >= 0) {
            pred = (int)(((Q02 << 7) + num) / (Q02 << 8));
            if (Al > 0 && pred >= (1 << Al))
@@ -599,14 +701,96 @@ decompress_smooth_data(j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
          }
          workspace[2] = (JCOEF)pred;
        }
+        if (change_dc) {
+          /* AC03 */
+          if ((Al = coef_bits[6]) != 0 && workspace[3] == 0) {
+            num = Q00 * (DC07 - DC09 + 2 * DC12 - 2 * DC14 + DC17 - DC19);
+            if (num >= 0) {
+              pred = (int)(((Q03 << 7) + num) / (Q03 << 8));
+              if (Al > 0 && pred >= (1 << Al))
+                pred = (1 << Al) - 1;
+            } else {
+              pred = (int)(((Q03 << 7) - num) / (Q03 << 8));
+              if (Al > 0 && pred >= (1 << Al))
+                pred = (1 << Al) - 1;
+              pred = -pred;
+            }
+            workspace[3] = (JCOEF)pred;
+          }
+          /* AC12 */
+          if ((Al = coef_bits[7]) != 0 && workspace[10] == 0) {
+            num = Q00 * (DC07 - 3 * DC08 + DC09 - DC17 + 3 * DC18 - DC19);
+            if (num >= 0) {
+              pred = (int)(((Q12 << 7) + num) / (Q12 << 8));
+              if (Al > 0 && pred >= (1 << Al))
+                pred = (1 << Al) - 1;
+            } else {
+              pred = (int)(((Q12 << 7) - num) / (Q12 << 8));
+              if (Al > 0 && pred >= (1 << Al))
+                pred = (1 << Al) - 1;
+              pred = -pred;
+            }
+            workspace[10] = (JCOEF)pred;
+          }
+          /* AC21 */
+          if ((Al = coef_bits[8]) != 0 && workspace[17] == 0) {
+            num = Q00 * (DC07 - DC09 - 3 * DC12 + 3 * DC14 + DC17 - DC19);
+            if (num >= 0) {
+              pred = (int)(((Q21 << 7) + num) / (Q21 << 8));
+              if (Al > 0 && pred >= (1 << Al))
+                pred = (1 << Al) - 1;
+            } else {
+              pred = (int)(((Q21 << 7) - num) / (Q21 << 8));
+              if (Al > 0 && pred >= (1 << Al))
+                pred = (1 << Al) - 1;
+              pred = -pred;
+            }
+            workspace[17] = (JCOEF)pred;
+          }
+          /* AC30 */
+          if ((Al = coef_bits[9]) != 0 && workspace[24] == 0) {
+            num = Q00 * (DC07 + 2 * DC08 + DC09 - DC17 - 2 * DC18 - DC19);
+            if (num >= 0) {
+              pred = (int)(((Q30 << 7) + num) / (Q30 << 8));
+              if (Al > 0 && pred >= (1 << Al))
+                pred = (1 << Al) - 1;
+            } else {
+              pred = (int)(((Q30 << 7) - num) / (Q30 << 8));
+              if (Al > 0 && pred >= (1 << Al))
+                pred = (1 << Al) - 1;
+              pred = -pred;
+            }
+            workspace[24] = (JCOEF)pred;
+          }
+          /* coef_bits[0] is non-negative.  Otherwise this function would not
+           * be called.
+           */
+          num = Q00 *
+                (-2 * DC01 - 6 * DC02 - 8 * DC03 - 6 * DC04 - 2 * DC05 -
+                 6 * DC06 + 6 * DC07 + 42 * DC08 + 6 * DC09 - 6 * DC10 -
+                 8 * DC11 + 42 * DC12 + 152 * DC13 + 42 * DC14 - 8 * DC15 -
+                 6 * DC16 + 6 * DC17 + 42 * DC18 + 6 * DC19 - 6 * DC20 -
+                 2 * DC21 - 6 * DC22 - 8 * DC23 - 6 * DC24 - 2 * DC25);
+          if (num >= 0) {
+            pred = (int)(((Q00 << 7) + num) / (Q00 << 8));
+          } else {
+            pred = (int)(((Q00 << 7) - num) / (Q00 << 8));
+            pred = -pred;
+          }
+          workspace[0] = (JCOEF)pred;
+        }  /* change_dc */
+
        /* OK, do the IDCT */
        (*inverse_DCT) (cinfo, compptr, (JCOEFPTR)workspace, output_ptr,
                        output_col);
        /* Advance for next column */
-        DC1 = DC2;  DC2 = DC3;
-        DC4 = DC5;  DC5 = DC6;
-        DC7 = DC8;  DC8 = DC9;
-        buffer_ptr++, prev_block_row++, next_block_row++;
+        DC01 = DC02;  DC02 = DC03;  DC03 = DC04;  DC04 = DC05;
+        DC06 = DC07;  DC07 = DC08;  DC08 = DC09;  DC09 = DC10;
+        DC11 = DC12;  DC12 = DC13;  DC13 = DC14;  DC14 = DC15;
+        DC16 = DC17;  DC17 = DC18;  DC18 = DC19;  DC19 = DC20;
+        DC21 = DC22;  DC22 = DC23;  DC23 = DC24;  DC24 = DC25;
+        buffer_ptr++, prev_block_row++, next_block_row++,
+          prev_prev_block_row++, next_next_block_row++;
        output_col += compptr->_DCT_scaled_size;
      }
      output_ptr += compptr->_DCT_scaled_size;
@@ -655,7 +839,7 @@ jinit_d_coef_controller(j_decompress_ptr cinfo, boolean need_full_buffer)
 #ifdef BLOCK_SMOOTHING_SUPPORTED
      /* If block smoothing could be used, need a bigger window */
      if (cinfo->progressive_mode)
-        access_rows *= 3;
+        access_rows *= 5;
 #endif
      coef->whole_image[ci] = (*cinfo->mem->request_virt_barray)
        ((j_common_ptr)cinfo, JPOOL_IMAGE, TRUE,
@@ -5,6 +5,7 @@
 * Copyright (C) 1994-1997, Thomas G. Lane.
 * libjpeg-turbo Modifications:
 * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright (C) 2020, Google, Inc.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 */
@@ -51,7 +52,7 @@ typedef struct {
 #ifdef BLOCK_SMOOTHING_SUPPORTED
  /* When doing block smoothing, we latch coefficient Al values here */
  int *coef_bits_latch;
-#define SAVED_COEFS  6          /* we save coef_bits[0..5] */
+#define SAVED_COEFS  10         /* we save coef_bits[0..9] */
 #endif
 } my_coef_controller;

@@ -45,9 +45,9 @@ ycc_rgb565_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
    outptr = *output_buf++;

    if (PACK_NEED_ALIGNMENT(outptr)) {
-      y  = GETJSAMPLE(*inptr0++);
-      cb = GETJSAMPLE(*inptr1++);
-      cr = GETJSAMPLE(*inptr2++);
+      y  = *inptr0++;
+      cb = *inptr1++;
+      cr = *inptr2++;
      r = range_limit[y + Crrtab[cr]];
      g = range_limit[y + ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
                                            SCALEBITS))];
@@ -58,18 +58,18 @@ ycc_rgb565_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
      num_cols--;
    }
    for (col = 0; col < (num_cols >> 1); col++) {
-      y  = GETJSAMPLE(*inptr0++);
-      cb = GETJSAMPLE(*inptr1++);
-      cr = GETJSAMPLE(*inptr2++);
+      y  = *inptr0++;
+      cb = *inptr1++;
+      cr = *inptr2++;
      r = range_limit[y + Crrtab[cr]];
      g = range_limit[y + ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
                                            SCALEBITS))];
      b = range_limit[y + Cbbtab[cb]];
      rgb = PACK_SHORT_565(r, g, b);

-      y  = GETJSAMPLE(*inptr0++);
-      cb = GETJSAMPLE(*inptr1++);
-      cr = GETJSAMPLE(*inptr2++);
+      y  = *inptr0++;
+      cb = *inptr1++;
+      cr = *inptr2++;
      r = range_limit[y + Crrtab[cr]];
      g = range_limit[y + ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
                                            SCALEBITS))];
@@ -80,9 +80,9 @@ ycc_rgb565_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
      outptr += 4;
    }
    if (num_cols & 1) {
-      y  = GETJSAMPLE(*inptr0);
-      cb = GETJSAMPLE(*inptr1);
-      cr = GETJSAMPLE(*inptr2);
+      y  = *inptr0;
+      cb = *inptr1;
+      cr = *inptr2;
      r = range_limit[y + Crrtab[cr]];
      g = range_limit[y + ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
                                            SCALEBITS))];
@@ -125,9 +125,9 @@ ycc_rgb565D_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
    input_row++;
    outptr = *output_buf++;
    if (PACK_NEED_ALIGNMENT(outptr)) {
-      y  = GETJSAMPLE(*inptr0++);
-      cb = GETJSAMPLE(*inptr1++);
-      cr = GETJSAMPLE(*inptr2++);
+      y  = *inptr0++;
+      cb = *inptr1++;
+      cr = *inptr2++;
      r = range_limit[DITHER_565_R(y + Crrtab[cr], d0)];
      g = range_limit[DITHER_565_G(y +
                                   ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
@@ -139,9 +139,9 @@ ycc_rgb565D_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
      num_cols--;
    }
    for (col = 0; col < (num_cols >> 1); col++) {
-      y  = GETJSAMPLE(*inptr0++);
-      cb = GETJSAMPLE(*inptr1++);
-      cr = GETJSAMPLE(*inptr2++);
+      y  = *inptr0++;
+      cb = *inptr1++;
+      cr = *inptr2++;
      r = range_limit[DITHER_565_R(y + Crrtab[cr], d0)];
      g = range_limit[DITHER_565_G(y +
                                   ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
@@ -150,9 +150,9 @@ ycc_rgb565D_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
      d0 = DITHER_ROTATE(d0);
      rgb = PACK_SHORT_565(r, g, b);

-      y  = GETJSAMPLE(*inptr0++);
-      cb = GETJSAMPLE(*inptr1++);
-      cr = GETJSAMPLE(*inptr2++);
+      y  = *inptr0++;
+      cb = *inptr1++;
+      cr = *inptr2++;
      r = range_limit[DITHER_565_R(y + Crrtab[cr], d0)];
      g = range_limit[DITHER_565_G(y +
                                   ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
@@ -165,9 +165,9 @@ ycc_rgb565D_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
      outptr += 4;
    }
    if (num_cols & 1) {
-      y  = GETJSAMPLE(*inptr0);
-      cb = GETJSAMPLE(*inptr1);
-      cr = GETJSAMPLE(*inptr2);
+      y  = *inptr0;
+      cb = *inptr1;
+      cr = *inptr2;
      r = range_limit[DITHER_565_R(y + Crrtab[cr], d0)];
      g = range_limit[DITHER_565_G(y +
                                   ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
@@ -202,32 +202,32 @@ rgb_rgb565_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
    input_row++;
    outptr = *output_buf++;
    if (PACK_NEED_ALIGNMENT(outptr)) {
-      r = GETJSAMPLE(*inptr0++);
-      g = GETJSAMPLE(*inptr1++);
-      b = GETJSAMPLE(*inptr2++);
+      r = *inptr0++;
+      g = *inptr1++;
+      b = *inptr2++;
      rgb = PACK_SHORT_565(r, g, b);
      *(INT16 *)outptr = (INT16)rgb;
      outptr += 2;
      num_cols--;
    }
    for (col = 0; col < (num_cols >> 1); col++) {
-      r = GETJSAMPLE(*inptr0++);
-      g = GETJSAMPLE(*inptr1++);
-      b = GETJSAMPLE(*inptr2++);
+      r = *inptr0++;
+      g = *inptr1++;
+      b = *inptr2++;
      rgb = PACK_SHORT_565(r, g, b);

-      r = GETJSAMPLE(*inptr0++);
-      g = GETJSAMPLE(*inptr1++);
-      b = GETJSAMPLE(*inptr2++);
+      r = *inptr0++;
+      g = *inptr1++;
+      b = *inptr2++;
      rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));

      WRITE_TWO_ALIGNED_PIXELS(outptr, rgb);
      outptr += 4;
    }
    if (num_cols & 1) {
-      r = GETJSAMPLE(*inptr0);
-      g = GETJSAMPLE(*inptr1);
-      b = GETJSAMPLE(*inptr2);
+      r = *inptr0;
+      g = *inptr1;
+      b = *inptr2;
      rgb = PACK_SHORT_565(r, g, b);
      *(INT16 *)outptr = (INT16)rgb;
    }
@@ -259,24 +259,24 @@ rgb_rgb565D_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
    input_row++;
    outptr = *output_buf++;
    if (PACK_NEED_ALIGNMENT(outptr)) {
-      r = range_limit[DITHER_565_R(GETJSAMPLE(*inptr0++), d0)];
-      g = range_limit[DITHER_565_G(GETJSAMPLE(*inptr1++), d0)];
-      b = range_limit[DITHER_565_B(GETJSAMPLE(*inptr2++), d0)];
+      r = range_limit[DITHER_565_R(*inptr0++, d0)];
+      g = range_limit[DITHER_565_G(*inptr1++, d0)];
+      b = range_limit[DITHER_565_B(*inptr2++, d0)];
      rgb = PACK_SHORT_565(r, g, b);
      *(INT16 *)outptr = (INT16)rgb;
      outptr += 2;
      num_cols--;
    }
    for (col = 0; col < (num_cols >> 1); col++) {
-      r = range_limit[DITHER_565_R(GETJSAMPLE(*inptr0++), d0)];
-      g = range_limit[DITHER_565_G(GETJSAMPLE(*inptr1++), d0)];
-      b = range_limit[DITHER_565_B(GETJSAMPLE(*inptr2++), d0)];
+      r = range_limit[DITHER_565_R(*inptr0++, d0)];
+      g = range_limit[DITHER_565_G(*inptr1++, d0)];
+      b = range_limit[DITHER_565_B(*inptr2++, d0)];
      d0 = DITHER_ROTATE(d0);
      rgb = PACK_SHORT_565(r, g, b);

-      r = range_limit[DITHER_565_R(GETJSAMPLE(*inptr0++), d0)];
-      g = range_limit[DITHER_565_G(GETJSAMPLE(*inptr1++), d0)];
-      b = range_limit[DITHER_565_B(GETJSAMPLE(*inptr2++), d0)];
+      r = range_limit[DITHER_565_R(*inptr0++, d0)];
+      g = range_limit[DITHER_565_G(*inptr1++, d0)];
+      b = range_limit[DITHER_565_B(*inptr2++, d0)];
      d0 = DITHER_ROTATE(d0);
      rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));

@@ -284,9 +284,9 @@ rgb_rgb565D_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
      outptr += 4;
    }
    if (num_cols & 1) {
-      r = range_limit[DITHER_565_R(GETJSAMPLE(*inptr0), d0)];
-      g = range_limit[DITHER_565_G(GETJSAMPLE(*inptr1), d0)];
-      b = range_limit[DITHER_565_B(GETJSAMPLE(*inptr2), d0)];
+      r = range_limit[DITHER_565_R(*inptr0, d0)];
+      g = range_limit[DITHER_565_G(*inptr1, d0)];
+      b = range_limit[DITHER_565_B(*inptr2, d0)];
      rgb = PACK_SHORT_565(r, g, b);
      *(INT16 *)outptr = (INT16)rgb;
    }
@@ -53,9 +53,9 @@ ycc_rgb_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
    input_row++;
    outptr = *output_buf++;
    for (col = 0; col < num_cols; col++) {
-      y  = GETJSAMPLE(inptr0[col]);
-      cb = GETJSAMPLE(inptr1[col]);
-      cr = GETJSAMPLE(inptr2[col]);
+      y  = inptr0[col];
+      cb = inptr1[col];
+      cr = inptr2[col];
      /* Range-limiting is essential due to noise introduced by DCT losses. */
      outptr[RGB_RED] =   range_limit[y + Crrtab[cr]];
      outptr[RGB_GREEN] = range_limit[y +
@@ -93,7 +93,6 @@ gray_rgb_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
    inptr = input_buf[0][input_row++];
    outptr = *output_buf++;
    for (col = 0; col < num_cols; col++) {
-      /* We can dispense with GETJSAMPLE() here */
      outptr[RGB_RED] = outptr[RGB_GREEN] = outptr[RGB_BLUE] = inptr[col];
      /* Set unused byte to 0xFF so it can be interpreted as an opaque */
      /* alpha channel value */
@@ -128,7 +127,6 @@ rgb_rgb_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
    input_row++;
    outptr = *output_buf++;
    for (col = 0; col < num_cols; col++) {
-      /* We can dispense with GETJSAMPLE() here */
      outptr[RGB_RED] = inptr0[col];
      outptr[RGB_GREEN] = inptr1[col];
      outptr[RGB_BLUE] = inptr2[col];
@@ -341,9 +341,9 @@ rgb_gray_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
    input_row++;
    outptr = *output_buf++;
    for (col = 0; col < num_cols; col++) {
-      r = GETJSAMPLE(inptr0[col]);
-      g = GETJSAMPLE(inptr1[col]);
-      b = GETJSAMPLE(inptr2[col]);
+      r = inptr0[col];
+      g = inptr1[col];
+      b = inptr2[col];
      /* Y */
      outptr[col] = (JSAMPLE)((ctab[r + R_Y_OFF] + ctab[g + G_Y_OFF] +
                               ctab[b + B_Y_OFF]) >> SCALEBITS);
@@ -550,9 +550,9 @@ ycck_cmyk_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
    input_row++;
    outptr = *output_buf++;
    for (col = 0; col < num_cols; col++) {
-      y  = GETJSAMPLE(inptr0[col]);
-      cb = GETJSAMPLE(inptr1[col]);
-      cr = GETJSAMPLE(inptr2[col]);
+      y  = inptr0[col];
+      cb = inptr1[col];
+      cr = inptr2[col];
      /* Range-limiting is essential due to noise introduced by DCT losses. */
      outptr[0] = range_limit[MAXJSAMPLE - (y + Crrtab[cr])];   /* red */
      outptr[1] = range_limit[MAXJSAMPLE - (y +                 /* green */
@@ -560,7 +560,7 @@ ycck_cmyk_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
                                                 SCALEBITS)))];
      outptr[2] = range_limit[MAXJSAMPLE - (y + Cbbtab[cb])];   /* blue */
      /* K passes through unchanged */
-      outptr[3] = inptr3[col];  /* don't need GETJSAMPLE here */
+      outptr[3] = inptr3[col];
      outptr += 4;
    }
  }
@@ -5,6 +5,7 @@
 * Copyright (C) 1991-1997, Thomas G. Lane.
 * libjpeg-turbo Modifications:
 * Copyright (C) 2009-2011, 2016, 2018-2019, D. R. Commander.
+ * Copyright (C) 2018, Matthias Räncker.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -39,24 +40,6 @@ typedef struct {
  int last_dc_val[MAX_COMPS_IN_SCAN]; /* last DC coef for each component */
 } savable_state;

-/* This macro is to work around compilers with missing or broken
- * structure assignment.  You'll need to fix this code if you have
- * such a compiler and you change MAX_COMPS_IN_SCAN.
- */
-
-#ifndef NO_STRUCT_ASSIGN
-#define ASSIGN_STATE(dest, src)  ((dest) = (src))
-#else
-#if MAX_COMPS_IN_SCAN == 4
-#define ASSIGN_STATE(dest, src) \
-  ((dest).last_dc_val[0] = (src).last_dc_val[0], \
-   (dest).last_dc_val[1] = (src).last_dc_val[1], \
-   (dest).last_dc_val[2] = (src).last_dc_val[2], \
-   (dest).last_dc_val[3] = (src).last_dc_val[3])
-#endif
-#endif
-
-
 typedef struct {
  struct jpeg_entropy_decoder pub; /* public fields */

@@ -325,7 +308,7 @@ jpeg_fill_bit_buffer(bitread_working_state *state,
        bytes_in_buffer = cinfo->src->bytes_in_buffer;
      }
      bytes_in_buffer--;
-      c = GETJOCTET(*next_input_byte++);
+      c = *next_input_byte++;

      /* If it's 0xFF, check and discard stuffed zero byte */
      if (c == 0xFF) {
@@ -342,7 +325,7 @@ jpeg_fill_bit_buffer(bitread_working_state *state,
            bytes_in_buffer = cinfo->src->bytes_in_buffer;
          }
          bytes_in_buffer--;
-          c = GETJOCTET(*next_input_byte++);
+          c = *next_input_byte++;
        } while (c == 0xFF);

        if (c == 0) {
@@ -405,8 +388,8 @@ no_more_bytes:

 #define GET_BYTE { \
  register int c0, c1; \
-  c0 = GETJOCTET(*buffer++); \
-  c1 = GETJOCTET(*buffer); \
+  c0 = *buffer++; \
+  c1 = *buffer; \
  /* Pre-execute most common case */ \
  get_buffer = (get_buffer << 8) | c0; \
  bits_left += 8; \
@@ -423,7 +406,7 @@ no_more_bytes:
  } \
 }

-#if SIZEOF_SIZE_T == 8 || defined(_WIN64)
+#if SIZEOF_SIZE_T == 8 || defined(_WIN64) || (defined(__x86_64__) && defined(__ILP32__))

 /* Pre-fetch 48 bytes, because the holding register is 64-bit */
 #define FILL_BIT_BUFFER_FAST \
@@ -557,6 +540,12 @@ process_restart(j_decompress_ptr cinfo)
 }


+#if defined(__has_feature)
+#if __has_feature(undefined_behavior_sanitizer)
+__attribute__((no_sanitize("signed-integer-overflow"),
+               no_sanitize("unsigned-integer-overflow")))
+#endif
+#endif
 LOCAL(boolean)
 decode_mcu_slow(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
 {
@@ -568,7 +557,7 @@ decode_mcu_slow(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)

  /* Load up working state */
  BITREAD_LOAD_STATE(cinfo, entropy->bitstate);
-  ASSIGN_STATE(state, entropy->saved);
+  state = entropy->saved;

  for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
    JBLOCKROW block = MCU_data ? MCU_data[blkn] : NULL;
@@ -589,11 +578,15 @@ decode_mcu_slow(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
    if (entropy->dc_needed[blkn]) {
      /* Convert DC difference to actual value, update last_dc_val */
      int ci = cinfo->MCU_membership[blkn];
-      /* This is really just
-       *   s += state.last_dc_val[ci];
-       * It is written this way in order to shut up UBSan.
+      /* Certain malformed JPEG images produce repeated DC coefficient
+       * differences of 2047 or -2047, which causes state.last_dc_val[ci] to
+       * grow until it overflows or underflows a 32-bit signed integer.  This
+       * behavior is, to the best of our understanding, innocuous, and it is
+       * unclear how to work around it without potentially affecting
+       * performance.  Thus, we (hopefully temporarily) suppress UBSan integer
+       * overflow errors for this function and decode_mcu_fast().
       */
-      s = (int)((unsigned int)s + (unsigned int)state.last_dc_val[ci]);
+      s += state.last_dc_val[ci];
      state.last_dc_val[ci] = s;
      if (block) {
        /* Output the DC coefficient (assumes jpeg_natural_order[0] = 0) */
@@ -653,11 +646,17 @@ decode_mcu_slow(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)

  /* Completed MCU, so update state */
  BITREAD_SAVE_STATE(cinfo, entropy->bitstate);
-  ASSIGN_STATE(entropy->saved, state);
+  entropy->saved = state;
  return TRUE;
 }


+#if defined(__has_feature)
+#if __has_feature(undefined_behavior_sanitizer)
+__attribute__((no_sanitize("signed-integer-overflow"),
+               no_sanitize("unsigned-integer-overflow")))
+#endif
+#endif
 LOCAL(boolean)
 decode_mcu_fast(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
 {
@@ -671,7 +670,7 @@ decode_mcu_fast(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
  /* Load up working state */
  BITREAD_LOAD_STATE(cinfo, entropy->bitstate);
  buffer = (JOCTET *)br_state.next_input_byte;
-  ASSIGN_STATE(state, entropy->saved);
+  state = entropy->saved;

  for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
    JBLOCKROW block = MCU_data ? MCU_data[blkn] : NULL;
@@ -688,7 +687,10 @@ decode_mcu_fast(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)

    if (entropy->dc_needed[blkn]) {
      int ci = cinfo->MCU_membership[blkn];
-      s = (int)((unsigned int)s + (unsigned int)state.last_dc_val[ci]);
+      /* Refer to the comment in decode_mcu_slow() regarding the supression of
+       * a UBSan integer overflow error in this line of code.
+       */
+      s += state.last_dc_val[ci];
      state.last_dc_val[ci] = s;
      if (block)
        (*block)[0] = (JCOEF)s;
@@ -740,7 +742,7 @@ decode_mcu_fast(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
  br_state.bytes_in_buffer -= (buffer - br_state.next_input_byte);
  br_state.next_input_byte = buffer;
  BITREAD_SAVE_STATE(cinfo, entropy->bitstate);
-  ASSIGN_STATE(entropy->saved, state);
+  entropy->saved = state;
  return TRUE;
 }

@@ -795,7 +797,8 @@ use_slow:
  }

  /* Account for restart interval (no-op if not using restarts) */
-  entropy->restarts_to_go--;
+  if (cinfo->restart_interval)
+    entropy->restarts_to_go--;

  return TRUE;
 }
@@ -4,7 +4,8 @@
 * This file was part of the Independent JPEG Group's software:
 * Copyright (C) 1991-1997, Thomas G. Lane.
 * libjpeg-turbo Modifications:
- * Copyright (C) 2010-2011, 2015-2016, D. R. Commander.
+ * Copyright (C) 2010-2011, 2015-2016, 2021, D. R. Commander.
+ * Copyright (C) 2018, Matthias Räncker.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -78,6 +79,11 @@ EXTERN(void) jpeg_make_d_derived_tbl(j_decompress_ptr cinfo, boolean isDC,
 typedef size_t bit_buf_type;            /* type of bit-extraction buffer */
 #define BIT_BUF_SIZE  64                /* size of buffer in bits */

+#elif defined(__x86_64__) && defined(__ILP32__)
+
+typedef unsigned long long bit_buf_type; /* type of bit-extraction buffer */
+#define BIT_BUF_SIZE  64                 /* size of buffer in bits */
+
 #else

 typedef unsigned long bit_buf_type;     /* type of bit-extraction buffer */
@@ -228,7 +234,10 @@ slowlabel: \
      s |= GET_BITS(1); \
      nb++; \
    } \
-    s = htbl->pub->huffval[(int)(s + htbl->valoffset[nb]) & 0xFF]; \
+    if (nb > 16) \
+      s = 0; \
+    else \
+      s = htbl->pub->huffval[(int)(s + htbl->valoffset[nb]) & 0xFF]; \
  }

 /* Out-of-line case for Huffman code fetching */
@@ -38,18 +38,18 @@ marker_is_icc(jpeg_saved_marker_ptr marker)
    marker->marker == ICC_MARKER &&
    marker->data_length >= ICC_OVERHEAD_LEN &&
    /* verify the identifying string */
-    GETJOCTET(marker->data[0]) == 0x49 &&
-    GETJOCTET(marker->data[1]) == 0x43 &&
-    GETJOCTET(marker->data[2]) == 0x43 &&
-    GETJOCTET(marker->data[3]) == 0x5F &&
-    GETJOCTET(marker->data[4]) == 0x50 &&
-    GETJOCTET(marker->data[5]) == 0x52 &&
-    GETJOCTET(marker->data[6]) == 0x4F &&
-    GETJOCTET(marker->data[7]) == 0x46 &&
-    GETJOCTET(marker->data[8]) == 0x49 &&
-    GETJOCTET(marker->data[9]) == 0x4C &&
-    GETJOCTET(marker->data[10]) == 0x45 &&
-    GETJOCTET(marker->data[11]) == 0x0;
+    marker->data[0] == 0x49 &&
+    marker->data[1] == 0x43 &&
+    marker->data[2] == 0x43 &&
+    marker->data[3] == 0x5F &&
+    marker->data[4] == 0x50 &&
+    marker->data[5] == 0x52 &&
+    marker->data[6] == 0x4F &&
+    marker->data[7] == 0x46 &&
+    marker->data[8] == 0x49 &&
+    marker->data[9] == 0x4C &&
+    marker->data[10] == 0x45 &&
+    marker->data[11] == 0x0;
 }


@@ -102,12 +102,12 @@ jpeg_read_icc_profile(j_decompress_ptr cinfo, JOCTET **icc_data_ptr,
  for (marker = cinfo->marker_list; marker != NULL; marker = marker->next) {
    if (marker_is_icc(marker)) {
      if (num_markers == 0)
-        num_markers = GETJOCTET(marker->data[13]);
-      else if (num_markers != GETJOCTET(marker->data[13])) {
+        num_markers = marker->data[13];
+      else if (num_markers != marker->data[13]) {
        WARNMS(cinfo, JWRN_BOGUS_ICC);  /* inconsistent num_markers fields */
        return FALSE;
      }
-      seq_no = GETJOCTET(marker->data[12]);
+      seq_no = marker->data[12];
      if (seq_no <= 0 || seq_no > num_markers) {
        WARNMS(cinfo, JWRN_BOGUS_ICC);  /* bogus sequence number */
        return FALSE;
@@ -154,7 +154,7 @@ jpeg_read_icc_profile(j_decompress_ptr cinfo, JOCTET **icc_data_ptr,
      JOCTET FAR *src_ptr;
      JOCTET *dst_ptr;
      unsigned int length;
-      seq_no = GETJOCTET(marker->data[12]);
+      seq_no = marker->data[12];
      dst_ptr = icc_data + data_offset[seq_no];
      src_ptr = marker->data + ICC_OVERHEAD_LEN;
      length = data_length[seq_no];
@@ -18,6 +18,7 @@

 #include "jinclude.h"
 #include "jdmainct.h"
+#include "jconfigint.h"


 /*
@@ -360,7 +361,7 @@ process_data_context_main(j_decompress_ptr cinfo, JSAMPARRAY output_buf,
    main_ptr->context_state = CTX_PREPARE_FOR_IMCU;
    if (*out_row_ctr >= out_rows_avail)
      return;                   /* Postprocessor exactly filled output buf */
-    /*FALLTHROUGH*/
+    FALLTHROUGH                 /*FALLTHROUGH*/
  case CTX_PREPARE_FOR_IMCU:
    /* Prepare to process first M-1 row groups of this iMCU row */
    main_ptr->rowgroup_ctr = 0;
@@ -371,7 +372,7 @@ process_data_context_main(j_decompress_ptr cinfo, JSAMPARRAY output_buf,
    if (main_ptr->iMCU_row_ctr == cinfo->total_iMCU_rows)
      set_bottom_pointers(cinfo);
    main_ptr->context_state = CTX_PROCESS_IMCU;
-    /*FALLTHROUGH*/
+    FALLTHROUGH                 /*FALLTHROUGH*/
  case CTX_PROCESS_IMCU:
    /* Call postprocessor using previously set pointers */
    (*cinfo->post->post_process_data) (cinfo,
@@ -151,7 +151,7 @@ typedef my_marker_reader *my_marker_ptr;
 #define INPUT_BYTE(cinfo, V, action) \
  MAKESTMT( MAKE_BYTE_AVAIL(cinfo, action); \
            bytes_in_buffer--; \
-            V = GETJOCTET(*next_input_byte++); )
+            V = *next_input_byte++; )

 /* As above, but read two bytes interpreted as an unsigned 16-bit integer.
 * V should be declared unsigned int or perhaps JLONG.
@@ -159,10 +159,10 @@ typedef my_marker_reader *my_marker_ptr;
 #define INPUT_2BYTES(cinfo, V, action) \
  MAKESTMT( MAKE_BYTE_AVAIL(cinfo, action); \
            bytes_in_buffer--; \
-            V = ((unsigned int)GETJOCTET(*next_input_byte++)) << 8; \
+            V = ((unsigned int)(*next_input_byte++)) << 8; \
            MAKE_BYTE_AVAIL(cinfo, action); \
            bytes_in_buffer--; \
-            V += GETJOCTET(*next_input_byte++); )
+            V += *next_input_byte++; )


 /*
@@ -608,18 +608,18 @@ examine_app0(j_decompress_ptr cinfo, JOCTET *data, unsigned int datalen,
  JLONG totallen = (JLONG)datalen + remaining;

  if (datalen >= APP0_DATA_LEN &&
-      GETJOCTET(data[0]) == 0x4A &&
-      GETJOCTET(data[1]) == 0x46 &&
-      GETJOCTET(data[2]) == 0x49 &&
-      GETJOCTET(data[3]) == 0x46 &&
-      GETJOCTET(data[4]) == 0) {
+      data[0] == 0x4A &&
+      data[1] == 0x46 &&
+      data[2] == 0x49 &&
+      data[3] == 0x46 &&
+      data[4] == 0) {
    /* Found JFIF APP0 marker: save info */
    cinfo->saw_JFIF_marker = TRUE;
-    cinfo->JFIF_major_version = GETJOCTET(data[5]);
-    cinfo->JFIF_minor_version = GETJOCTET(data[6]);
-    cinfo->density_unit = GETJOCTET(data[7]);
-    cinfo->X_density = (GETJOCTET(data[8]) << 8) + GETJOCTET(data[9]);
-    cinfo->Y_density = (GETJOCTET(data[10]) << 8) + GETJOCTET(data[11]);
+    cinfo->JFIF_major_version = data[5];
+    cinfo->JFIF_minor_version = data[6];
+    cinfo->density_unit = data[7];
+    cinfo->X_density = (data[8] << 8) + data[9];
+    cinfo->Y_density = (data[10] << 8) + data[11];
    /* Check version.
     * Major version must be 1, anything else signals an incompatible change.
     * (We used to treat this as an error, but now it's a nonfatal warning,
@@ -634,24 +634,22 @@ examine_app0(j_decompress_ptr cinfo, JOCTET *data, unsigned int datalen,
             cinfo->JFIF_major_version, cinfo->JFIF_minor_version,
             cinfo->X_density, cinfo->Y_density, cinfo->density_unit);
    /* Validate thumbnail dimensions and issue appropriate messages */
-    if (GETJOCTET(data[12]) | GETJOCTET(data[13]))
-      TRACEMS2(cinfo, 1, JTRC_JFIF_THUMBNAIL,
-               GETJOCTET(data[12]), GETJOCTET(data[13]));
+    if (data[12] | data[13])
+      TRACEMS2(cinfo, 1, JTRC_JFIF_THUMBNAIL, data[12], data[13]);
    totallen -= APP0_DATA_LEN;
-    if (totallen !=
-        ((JLONG)GETJOCTET(data[12]) * (JLONG)GETJOCTET(data[13]) * (JLONG)3))
+    if (totallen != ((JLONG)data[12] * (JLONG)data[13] * (JLONG)3))
      TRACEMS1(cinfo, 1, JTRC_JFIF_BADTHUMBNAILSIZE, (int)totallen);
  } else if (datalen >= 6 &&
-             GETJOCTET(data[0]) == 0x4A &&
-             GETJOCTET(data[1]) == 0x46 &&
-             GETJOCTET(data[2]) == 0x58 &&
-             GETJOCTET(data[3]) == 0x58 &&
-             GETJOCTET(data[4]) == 0) {
+             data[0] == 0x4A &&
+             data[1] == 0x46 &&
+             data[2] == 0x58 &&
+             data[3] == 0x58 &&
+             data[4] == 0) {
    /* Found JFIF "JFXX" extension APP0 marker */
    /* The library doesn't actually do anything with these,
     * but we try to produce a helpful trace message.
     */
-    switch (GETJOCTET(data[5])) {
+    switch (data[5]) {
    case 0x10:
      TRACEMS1(cinfo, 1, JTRC_THUMB_JPEG, (int)totallen);
      break;
@@ -662,8 +660,7 @@ examine_app0(j_decompress_ptr cinfo, JOCTET *data, unsigned int datalen,
      TRACEMS1(cinfo, 1, JTRC_THUMB_RGB, (int)totallen);
      break;
    default:
-      TRACEMS2(cinfo, 1, JTRC_JFIF_EXTENSION,
-               GETJOCTET(data[5]), (int)totallen);
+      TRACEMS2(cinfo, 1, JTRC_JFIF_EXTENSION, data[5], (int)totallen);
      break;
    }
  } else {
@@ -684,16 +681,16 @@ examine_app14(j_decompress_ptr cinfo, JOCTET *data, unsigned int datalen,
  unsigned int version, flags0, flags1, transform;

  if (datalen >= APP14_DATA_LEN &&
-      GETJOCTET(data[0]) == 0x41 &&
-      GETJOCTET(data[1]) == 0x64 &&
-      GETJOCTET(data[2]) == 0x6F &&
-      GETJOCTET(data[3]) == 0x62 &&
-      GETJOCTET(data[4]) == 0x65) {
+      data[0] == 0x41 &&
+      data[1] == 0x64 &&
+      data[2] == 0x6F &&
+      data[3] == 0x62 &&
+      data[4] == 0x65) {
    /* Found Adobe APP14 marker */
-    version = (GETJOCTET(data[5]) << 8) + GETJOCTET(data[6]);
-    flags0 = (GETJOCTET(data[7]) << 8) + GETJOCTET(data[8]);
-    flags1 = (GETJOCTET(data[9]) << 8) + GETJOCTET(data[10]);
-    transform = GETJOCTET(data[11]);
+    version = (data[5] << 8) + data[6];
+    flags0 = (data[7] << 8) + data[8];
+    flags1 = (data[9] << 8) + data[10];
+    transform = data[11];
    TRACEMS4(cinfo, 1, JTRC_ADOBE, version, flags0, flags1, transform);
    cinfo->saw_Adobe_marker = TRUE;
    cinfo->Adobe_transform = (UINT8)transform;
@@ -5,7 +5,7 @@
 * Copyright (C) 1991-1997, Thomas G. Lane.
 * Modified 2002-2009 by Guido Vollbeding.
 * libjpeg-turbo Modifications:
- * Copyright (C) 2009-2011, 2016, D. R. Commander.
+ * Copyright (C) 2009-2011, 2016, 2019, D. R. Commander.
 * Copyright (C) 2013, Linaro Limited.
 * Copyright (C) 2015, Google, Inc.
 * For conditions of distribution and use, see the accompanying README.ijg
@@ -22,7 +22,6 @@
 #include "jpeglib.h"
 #include "jpegcomp.h"
 #include "jdmaster.h"
-#include "jsimd.h"


 /*
@@ -70,17 +69,6 @@ use_merged_upsample(j_decompress_ptr cinfo)
      cinfo->comp_info[1]._DCT_scaled_size != cinfo->_min_DCT_scaled_size ||
      cinfo->comp_info[2]._DCT_scaled_size != cinfo->_min_DCT_scaled_size)
    return FALSE;
-#ifdef WITH_SIMD
-  /* If YCbCr-to-RGB color conversion is SIMD-accelerated but merged upsampling
-     isn't, then disabling merged upsampling is likely to be faster when
-     decompressing YCbCr JPEG images. */
-  if (!jsimd_can_h2v2_merged_upsample() && !jsimd_can_h2v1_merged_upsample() &&
-      jsimd_can_ycc_rgb() && cinfo->jpeg_color_space == JCS_YCbCr &&
-      (cinfo->out_color_space == JCS_RGB ||
-       (cinfo->out_color_space >= JCS_EXT_RGB &&
-        cinfo->out_color_space <= JCS_EXT_ARGB)))
-    return FALSE;
-#endif
  /* ??? also need to test for upsample-time rescaling, when & if supported */
  return TRUE;                  /* by golly, it'll work... */
 #else
@@ -580,6 +568,7 @@ master_selection(j_decompress_ptr cinfo)
   */
  cinfo->master->first_iMCU_col = 0;
  cinfo->master->last_iMCU_col = cinfo->MCUs_per_row - 1;
+  cinfo->master->last_good_iMCU_row = 0;

 #ifdef D_MULTISCAN_FILES_SUPPORTED
  /* If jpeg_start_decompress will read the whole file, initialize
@@ -43,20 +43,20 @@ h2v1_merged_upsample_565_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
  /* Loop for each pair of output pixels */
  for (col = cinfo->output_width >> 1; col > 0; col--) {
    /* Do the chroma part of the calculation */
-    cb = GETJSAMPLE(*inptr1++);
-    cr = GETJSAMPLE(*inptr2++);
+    cb = *inptr1++;
+    cr = *inptr2++;
    cred = Crrtab[cr];
    cgreen = (int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
    cblue = Cbbtab[cb];

    /* Fetch 2 Y values and emit 2 pixels */
-    y  = GETJSAMPLE(*inptr0++);
+    y  = *inptr0++;
    r = range_limit[y + cred];
    g = range_limit[y + cgreen];
    b = range_limit[y + cblue];
    rgb = PACK_SHORT_565(r, g, b);

-    y  = GETJSAMPLE(*inptr0++);
+    y  = *inptr0++;
    r = range_limit[y + cred];
    g = range_limit[y + cgreen];
    b = range_limit[y + cblue];
@@ -68,12 +68,12 @@ h2v1_merged_upsample_565_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,

  /* If image width is odd, do the last output column separately */
  if (cinfo->output_width & 1) {
-    cb = GETJSAMPLE(*inptr1);
-    cr = GETJSAMPLE(*inptr2);
+    cb = *inptr1;
+    cr = *inptr2;
    cred = Crrtab[cr];
    cgreen = (int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
    cblue = Cbbtab[cb];
-    y  = GETJSAMPLE(*inptr0);
+    y  = *inptr0;
    r = range_limit[y + cred];
    g = range_limit[y + cgreen];
    b = range_limit[y + cblue];
@@ -115,21 +115,21 @@ h2v1_merged_upsample_565D_internal(j_decompress_ptr cinfo,
  /* Loop for each pair of output pixels */
  for (col = cinfo->output_width >> 1; col > 0; col--) {
    /* Do the chroma part of the calculation */
-    cb = GETJSAMPLE(*inptr1++);
-    cr = GETJSAMPLE(*inptr2++);
+    cb = *inptr1++;
+    cr = *inptr2++;
    cred = Crrtab[cr];
    cgreen = (int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
    cblue = Cbbtab[cb];

    /* Fetch 2 Y values and emit 2 pixels */
-    y  = GETJSAMPLE(*inptr0++);
+    y  = *inptr0++;
    r = range_limit[DITHER_565_R(y + cred, d0)];
    g = range_limit[DITHER_565_G(y + cgreen, d0)];
    b = range_limit[DITHER_565_B(y + cblue, d0)];
    d0 = DITHER_ROTATE(d0);
    rgb = PACK_SHORT_565(r, g, b);

-    y  = GETJSAMPLE(*inptr0++);
+    y  = *inptr0++;
    r = range_limit[DITHER_565_R(y + cred, d0)];
    g = range_limit[DITHER_565_G(y + cgreen, d0)];
    b = range_limit[DITHER_565_B(y + cblue, d0)];
@@ -142,12 +142,12 @@ h2v1_merged_upsample_565D_internal(j_decompress_ptr cinfo,

  /* If image width is odd, do the last output column separately */
  if (cinfo->output_width & 1) {
-    cb = GETJSAMPLE(*inptr1);
-    cr = GETJSAMPLE(*inptr2);
+    cb = *inptr1;
+    cr = *inptr2;
    cred = Crrtab[cr];
    cgreen = (int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
    cblue = Cbbtab[cb];
-    y  = GETJSAMPLE(*inptr0);
+    y  = *inptr0;
    r = range_limit[DITHER_565_R(y + cred, d0)];
    g = range_limit[DITHER_565_G(y + cgreen, d0)];
    b = range_limit[DITHER_565_B(y + cblue, d0)];
@@ -189,20 +189,20 @@ h2v2_merged_upsample_565_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
  /* Loop for each group of output pixels */
  for (col = cinfo->output_width >> 1; col > 0; col--) {
    /* Do the chroma part of the calculation */
-    cb = GETJSAMPLE(*inptr1++);
-    cr = GETJSAMPLE(*inptr2++);
+    cb = *inptr1++;
+    cr = *inptr2++;
    cred = Crrtab[cr];
    cgreen = (int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
    cblue = Cbbtab[cb];

    /* Fetch 4 Y values and emit 4 pixels */
-    y  = GETJSAMPLE(*inptr00++);
+    y  = *inptr00++;
    r = range_limit[y + cred];
    g = range_limit[y + cgreen];
    b = range_limit[y + cblue];
    rgb = PACK_SHORT_565(r, g, b);

-    y  = GETJSAMPLE(*inptr00++);
+    y  = *inptr00++;
    r = range_limit[y + cred];
    g = range_limit[y + cgreen];
    b = range_limit[y + cblue];
@@ -211,13 +211,13 @@ h2v2_merged_upsample_565_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
    WRITE_TWO_PIXELS(outptr0, rgb);
    outptr0 += 4;

-    y  = GETJSAMPLE(*inptr01++);
+    y  = *inptr01++;
    r = range_limit[y + cred];
    g = range_limit[y + cgreen];
    b = range_limit[y + cblue];
    rgb = PACK_SHORT_565(r, g, b);

-    y  = GETJSAMPLE(*inptr01++);
+    y  = *inptr01++;
    r = range_limit[y + cred];
    g = range_limit[y + cgreen];
    b = range_limit[y + cblue];
@@ -229,20 +229,20 @@ h2v2_merged_upsample_565_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,

  /* If image width is odd, do the last output column separately */
  if (cinfo->output_width & 1) {
-    cb = GETJSAMPLE(*inptr1);
-    cr = GETJSAMPLE(*inptr2);
+    cb = *inptr1;
+    cr = *inptr2;
    cred = Crrtab[cr];
    cgreen = (int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
    cblue = Cbbtab[cb];

-    y  = GETJSAMPLE(*inptr00);
+    y  = *inptr00;
    r = range_limit[y + cred];
    g = range_limit[y + cgreen];
    b = range_limit[y + cblue];
    rgb = PACK_SHORT_565(r, g, b);
    *(INT16 *)outptr0 = (INT16)rgb;

-    y  = GETJSAMPLE(*inptr01);
+    y  = *inptr01;
    r = range_limit[y + cred];
    g = range_limit[y + cgreen];
    b = range_limit[y + cblue];
@@ -287,21 +287,21 @@ h2v2_merged_upsample_565D_internal(j_decompress_ptr cinfo,
  /* Loop for each group of output pixels */
  for (col = cinfo->output_width >> 1; col > 0; col--) {
    /* Do the chroma part of the calculation */
-    cb = GETJSAMPLE(*inptr1++);
-    cr = GETJSAMPLE(*inptr2++);
+    cb = *inptr1++;
+    cr = *inptr2++;
    cred = Crrtab[cr];
    cgreen = (int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
    cblue = Cbbtab[cb];

    /* Fetch 4 Y values and emit 4 pixels */
-    y  = GETJSAMPLE(*inptr00++);
+    y  = *inptr00++;
    r = range_limit[DITHER_565_R(y + cred, d0)];
    g = range_limit[DITHER_565_G(y + cgreen, d0)];
    b = range_limit[DITHER_565_B(y + cblue, d0)];
    d0 = DITHER_ROTATE(d0);
    rgb = PACK_SHORT_565(r, g, b);

-    y  = GETJSAMPLE(*inptr00++);
+    y  = *inptr00++;
    r = range_limit[DITHER_565_R(y + cred, d0)];
    g = range_limit[DITHER_565_G(y + cgreen, d0)];
    b = range_limit[DITHER_565_B(y + cblue, d0)];
@@ -311,14 +311,14 @@ h2v2_merged_upsample_565D_internal(j_decompress_ptr cinfo,
    WRITE_TWO_PIXELS(outptr0, rgb);
    outptr0 += 4;

-    y  = GETJSAMPLE(*inptr01++);
+    y  = *inptr01++;
    r = range_limit[DITHER_565_R(y + cred, d1)];
    g = range_limit[DITHER_565_G(y + cgreen, d1)];
    b = range_limit[DITHER_565_B(y + cblue, d1)];
    d1 = DITHER_ROTATE(d1);
    rgb = PACK_SHORT_565(r, g, b);

-    y  = GETJSAMPLE(*inptr01++);
+    y  = *inptr01++;
    r = range_limit[DITHER_565_R(y + cred, d1)];
    g = range_limit[DITHER_565_G(y + cgreen, d1)];
    b = range_limit[DITHER_565_B(y + cblue, d1)];
@@ -331,20 +331,20 @@ h2v2_merged_upsample_565D_internal(j_decompress_ptr cinfo,

  /* If image width is odd, do the last output column separately */
  if (cinfo->output_width & 1) {
-    cb = GETJSAMPLE(*inptr1);
-    cr = GETJSAMPLE(*inptr2);
+    cb = *inptr1;
+    cr = *inptr2;
    cred = Crrtab[cr];
    cgreen = (int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
    cblue = Cbbtab[cb];

-    y  = GETJSAMPLE(*inptr00);
+    y  = *inptr00;
    r = range_limit[DITHER_565_R(y + cred, d0)];
    g = range_limit[DITHER_565_G(y + cgreen, d0)];
    b = range_limit[DITHER_565_B(y + cblue, d0)];
    rgb = PACK_SHORT_565(r, g, b);
    *(INT16 *)outptr0 = (INT16)rgb;

-    y  = GETJSAMPLE(*inptr01);
+    y  = *inptr01;
    r = range_limit[DITHER_565_R(y + cred, d1)];
    g = range_limit[DITHER_565_G(y + cgreen, d1)];
    b = range_limit[DITHER_565_B(y + cblue, d1)];
@@ -46,13 +46,13 @@ h2v1_merged_upsample_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
  /* Loop for each pair of output pixels */
  for (col = cinfo->output_width >> 1; col > 0; col--) {
    /* Do the chroma part of the calculation */
-    cb = GETJSAMPLE(*inptr1++);
-    cr = GETJSAMPLE(*inptr2++);
+    cb = *inptr1++;
+    cr = *inptr2++;
    cred = Crrtab[cr];
    cgreen = (int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
    cblue = Cbbtab[cb];
    /* Fetch 2 Y values and emit 2 pixels */
-    y  = GETJSAMPLE(*inptr0++);
+    y  = *inptr0++;
    outptr[RGB_RED] =   range_limit[y + cred];
    outptr[RGB_GREEN] = range_limit[y + cgreen];
    outptr[RGB_BLUE] =  range_limit[y + cblue];
@@ -60,7 +60,7 @@ h2v1_merged_upsample_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
    outptr[RGB_ALPHA] = 0xFF;
 #endif
    outptr += RGB_PIXELSIZE;
-    y  = GETJSAMPLE(*inptr0++);
+    y  = *inptr0++;
    outptr[RGB_RED] =   range_limit[y + cred];
    outptr[RGB_GREEN] = range_limit[y + cgreen];
    outptr[RGB_BLUE] =  range_limit[y + cblue];
@@ -71,12 +71,12 @@ h2v1_merged_upsample_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
  }
  /* If image width is odd, do the last output column separately */
  if (cinfo->output_width & 1) {
-    cb = GETJSAMPLE(*inptr1);
-    cr = GETJSAMPLE(*inptr2);
+    cb = *inptr1;
+    cr = *inptr2;
    cred = Crrtab[cr];
    cgreen = (int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
    cblue = Cbbtab[cb];
-    y  = GETJSAMPLE(*inptr0);
+    y  = *inptr0;
    outptr[RGB_RED] =   range_limit[y + cred];
    outptr[RGB_GREEN] = range_limit[y + cgreen];
    outptr[RGB_BLUE] =  range_limit[y + cblue];
@@ -120,13 +120,13 @@ h2v2_merged_upsample_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
  /* Loop for each group of output pixels */
  for (col = cinfo->output_width >> 1; col > 0; col--) {
    /* Do the chroma part of the calculation */
-    cb = GETJSAMPLE(*inptr1++);
-    cr = GETJSAMPLE(*inptr2++);
+    cb = *inptr1++;
+    cr = *inptr2++;
    cred = Crrtab[cr];
    cgreen = (int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
    cblue = Cbbtab[cb];
    /* Fetch 4 Y values and emit 4 pixels */
-    y  = GETJSAMPLE(*inptr00++);
+    y  = *inptr00++;
    outptr0[RGB_RED] =   range_limit[y + cred];
    outptr0[RGB_GREEN] = range_limit[y + cgreen];
    outptr0[RGB_BLUE] =  range_limit[y + cblue];
@@ -134,7 +134,7 @@ h2v2_merged_upsample_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
    outptr0[RGB_ALPHA] = 0xFF;
 #endif
    outptr0 += RGB_PIXELSIZE;
-    y  = GETJSAMPLE(*inptr00++);
+    y  = *inptr00++;
    outptr0[RGB_RED] =   range_limit[y + cred];
    outptr0[RGB_GREEN] = range_limit[y + cgreen];
    outptr0[RGB_BLUE] =  range_limit[y + cblue];
@@ -142,7 +142,7 @@ h2v2_merged_upsample_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
    outptr0[RGB_ALPHA] = 0xFF;
 #endif
    outptr0 += RGB_PIXELSIZE;
-    y  = GETJSAMPLE(*inptr01++);
+    y  = *inptr01++;
    outptr1[RGB_RED] =   range_limit[y + cred];
    outptr1[RGB_GREEN] = range_limit[y + cgreen];
    outptr1[RGB_BLUE] =  range_limit[y + cblue];
@@ -150,7 +150,7 @@ h2v2_merged_upsample_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
    outptr1[RGB_ALPHA] = 0xFF;
 #endif
    outptr1 += RGB_PIXELSIZE;
-    y  = GETJSAMPLE(*inptr01++);
+    y  = *inptr01++;
    outptr1[RGB_RED] =   range_limit[y + cred];
    outptr1[RGB_GREEN] = range_limit[y + cgreen];
    outptr1[RGB_BLUE] =  range_limit[y + cblue];
@@ -161,19 +161,19 @@ h2v2_merged_upsample_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
  }
  /* If image width is odd, do the last output column separately */
  if (cinfo->output_width & 1) {
-    cb = GETJSAMPLE(*inptr1);
-    cr = GETJSAMPLE(*inptr2);
+    cb = *inptr1;
+    cr = *inptr2;
    cred = Crrtab[cr];
    cgreen = (int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
    cblue = Cbbtab[cb];
-    y  = GETJSAMPLE(*inptr00);
+    y  = *inptr00;
    outptr0[RGB_RED] =   range_limit[y + cred];
    outptr0[RGB_GREEN] = range_limit[y + cgreen];
    outptr0[RGB_BLUE] =  range_limit[y + cblue];
 #ifdef RGB_ALPHA
    outptr0[RGB_ALPHA] = 0xFF;
 #endif
-    y  = GETJSAMPLE(*inptr01);
+    y  = *inptr01;
    outptr1[RGB_RED] =   range_limit[y + cred];
    outptr1[RGB_GREEN] = range_limit[y + cgreen];
    outptr1[RGB_BLUE] =  range_limit[y + cblue];
@@ -4,7 +4,7 @@
 * This file was part of the Independent JPEG Group's software:
 * Copyright (C) 1995-1997, Thomas G. Lane.
 * libjpeg-turbo Modifications:
- * Copyright (C) 2015-2016, 2018, D. R. Commander.
+ * Copyright (C) 2015-2016, 2018-2021, D. R. Commander.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -41,25 +41,6 @@ typedef struct {
  int last_dc_val[MAX_COMPS_IN_SCAN];   /* last DC coef for each component */
 } savable_state;

-/* This macro is to work around compilers with missing or broken
- * structure assignment.  You'll need to fix this code if you have
- * such a compiler and you change MAX_COMPS_IN_SCAN.
- */
-
-#ifndef NO_STRUCT_ASSIGN
-#define ASSIGN_STATE(dest, src)  ((dest) = (src))
-#else
-#if MAX_COMPS_IN_SCAN == 4
-#define ASSIGN_STATE(dest, src) \
-  ((dest).EOBRUN = (src).EOBRUN, \
-   (dest).last_dc_val[0] = (src).last_dc_val[0], \
-   (dest).last_dc_val[1] = (src).last_dc_val[1], \
-   (dest).last_dc_val[2] = (src).last_dc_val[2], \
-   (dest).last_dc_val[3] = (src).last_dc_val[3])
-#endif
-#endif
-
-
 typedef struct {
  struct jpeg_entropy_decoder pub; /* public fields */

@@ -102,7 +83,7 @@ start_pass_phuff_decoder(j_decompress_ptr cinfo)
  boolean is_DC_band, bad;
  int ci, coefi, tbl;
  d_derived_tbl **pdtbl;
-  int *coef_bit_ptr;
+  int *coef_bit_ptr, *prev_coef_bit_ptr;
  jpeg_component_info *compptr;

  is_DC_band = (cinfo->Ss == 0);
@@ -143,8 +124,15 @@ start_pass_phuff_decoder(j_decompress_ptr cinfo)
  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
    int cindex = cinfo->cur_comp_info[ci]->component_index;
    coef_bit_ptr = &cinfo->coef_bits[cindex][0];
+    prev_coef_bit_ptr = &cinfo->coef_bits[cindex + cinfo->num_components][0];
    if (!is_DC_band && coef_bit_ptr[0] < 0) /* AC without prior DC scan */
      WARNMS2(cinfo, JWRN_BOGUS_PROGRESSION, cindex, 0);
+    for (coefi = MIN(cinfo->Ss, 1); coefi <= MAX(cinfo->Se, 9); coefi++) {
+      if (cinfo->input_scan_number > 1)
+        prev_coef_bit_ptr[coefi] = coef_bit_ptr[coefi];
+      else
+        prev_coef_bit_ptr[coefi] = 0;
+    }
    for (coefi = cinfo->Ss; coefi <= cinfo->Se; coefi++) {
      int expected = (coef_bit_ptr[coefi] < 0) ? 0 : coef_bit_ptr[coefi];
      if (cinfo->Ah != expected)
@@ -323,7 +311,7 @@ decode_mcu_DC_first(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)

    /* Load up working state */
    BITREAD_LOAD_STATE(cinfo, entropy->bitstate);
-    ASSIGN_STATE(state, entropy->saved);
+    state = entropy->saved;

    /* Outer loop handles each block in the MCU */

@@ -356,11 +344,12 @@ decode_mcu_DC_first(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)

    /* Completed MCU, so update state */
    BITREAD_SAVE_STATE(cinfo, entropy->bitstate);
-    ASSIGN_STATE(entropy->saved, state);
+    entropy->saved = state;
  }

  /* Account for restart interval (no-op if not using restarts) */
-  entropy->restarts_to_go--;
+  if (cinfo->restart_interval)
+    entropy->restarts_to_go--;

  return TRUE;
 }
@@ -444,7 +433,8 @@ decode_mcu_AC_first(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
  }

  /* Account for restart interval (no-op if not using restarts) */
-  entropy->restarts_to_go--;
+  if (cinfo->restart_interval)
+    entropy->restarts_to_go--;

  return TRUE;
 }
@@ -495,7 +485,8 @@ decode_mcu_DC_refine(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
  BITREAD_SAVE_STATE(cinfo, entropy->bitstate);

  /* Account for restart interval (no-op if not using restarts) */
-  entropy->restarts_to_go--;
+  if (cinfo->restart_interval)
+    entropy->restarts_to_go--;

  return TRUE;
 }
@@ -638,7 +629,8 @@ decode_mcu_AC_refine(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
  }

  /* Account for restart interval (no-op if not using restarts) */
-  entropy->restarts_to_go--;
+  if (cinfo->restart_interval)
+    entropy->restarts_to_go--;

  return TRUE;

@@ -676,7 +668,7 @@ jinit_phuff_decoder(j_decompress_ptr cinfo)
  /* Create progression status table */
  cinfo->coef_bits = (int (*)[DCTSIZE2])
    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
-                                cinfo->num_components * DCTSIZE2 *
+                                cinfo->num_components * 2 * DCTSIZE2 *
                                sizeof(int));
  coef_bit_ptr = &cinfo->coef_bits[0][0];
  for (ci = 0; ci < cinfo->num_components; ci++)
@@ -8,7 +8,7 @@
 * Copyright (C) 2010, 2015-2016, D. R. Commander.
 * Copyright (C) 2014, MIPS Technologies, Inc., California.
 * Copyright (C) 2015, Google, Inc.
- * Copyright (C) 2019, Arm Limited.
+ * Copyright (C) 2019-2020, Arm Limited.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -177,7 +177,7 @@ int_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
    outptr = output_data[outrow];
    outend = outptr + cinfo->output_width;
    while (outptr < outend) {
-      invalue = *inptr++;       /* don't need GETJSAMPLE() here */
+      invalue = *inptr++;
      for (h = h_expand; h > 0; h--) {
        *outptr++ = invalue;
      }
@@ -213,7 +213,7 @@ h2v1_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
    outptr = output_data[inrow];
    outend = outptr + cinfo->output_width;
    while (outptr < outend) {
-      invalue = *inptr++;       /* don't need GETJSAMPLE() here */
+      invalue = *inptr++;
      *outptr++ = invalue;
      *outptr++ = invalue;
    }
@@ -242,7 +242,7 @@ h2v2_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
    outptr = output_data[outrow];
    outend = outptr + cinfo->output_width;
    while (outptr < outend) {
-      invalue = *inptr++;       /* don't need GETJSAMPLE() here */
+      invalue = *inptr++;
      *outptr++ = invalue;
      *outptr++ = invalue;
    }
@@ -283,20 +283,20 @@ h2v1_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
    inptr = input_data[inrow];
    outptr = output_data[inrow];
    /* Special case for first column */
-    invalue = GETJSAMPLE(*inptr++);
+    invalue = *inptr++;
    *outptr++ = (JSAMPLE)invalue;
-    *outptr++ = (JSAMPLE)((invalue * 3 + GETJSAMPLE(*inptr) + 2) >> 2);
+    *outptr++ = (JSAMPLE)((invalue * 3 + inptr[0] + 2) >> 2);

    for (colctr = compptr->downsampled_width - 2; colctr > 0; colctr--) {
      /* General case: 3/4 * nearer pixel + 1/4 * further pixel */
-      invalue = GETJSAMPLE(*inptr++) * 3;
-      *outptr++ = (JSAMPLE)((invalue + GETJSAMPLE(inptr[-2]) + 1) >> 2);
-      *outptr++ = (JSAMPLE)((invalue + GETJSAMPLE(*inptr) + 2) >> 2);
+      invalue = (*inptr++) * 3;
+      *outptr++ = (JSAMPLE)((invalue + inptr[-2] + 1) >> 2);
+      *outptr++ = (JSAMPLE)((invalue + inptr[0] + 2) >> 2);
    }

    /* Special case for last column */
-    invalue = GETJSAMPLE(*inptr);
-    *outptr++ = (JSAMPLE)((invalue * 3 + GETJSAMPLE(inptr[-1]) + 1) >> 2);
+    invalue = *inptr;
+    *outptr++ = (JSAMPLE)((invalue * 3 + inptr[-1] + 1) >> 2);
    *outptr++ = (JSAMPLE)invalue;
  }
 }
@@ -338,7 +338,7 @@ h1v2_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
      outptr = output_data[outrow++];

      for (colctr = 0; colctr < compptr->downsampled_width; colctr++) {
-        thiscolsum = GETJSAMPLE(*inptr0++) * 3 + GETJSAMPLE(*inptr1++);
+        thiscolsum = (*inptr0++) * 3 + (*inptr1++);
        *outptr++ = (JSAMPLE)((thiscolsum + bias) >> 2);
      }
    }
@@ -381,8 +381,8 @@ h2v2_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
      outptr = output_data[outrow++];

      /* Special case for first column */
-      thiscolsum = GETJSAMPLE(*inptr0++) * 3 + GETJSAMPLE(*inptr1++);
-      nextcolsum = GETJSAMPLE(*inptr0++) * 3 + GETJSAMPLE(*inptr1++);
+      thiscolsum = (*inptr0++) * 3 + (*inptr1++);
+      nextcolsum = (*inptr0++) * 3 + (*inptr1++);
      *outptr++ = (JSAMPLE)((thiscolsum * 4 + 8) >> 4);
      *outptr++ = (JSAMPLE)((thiscolsum * 3 + nextcolsum + 7) >> 4);
      lastcolsum = thiscolsum;  thiscolsum = nextcolsum;
@@ -390,7 +390,7 @@ h2v2_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
      for (colctr = compptr->downsampled_width - 2; colctr > 0; colctr--) {
        /* General case: 3/4 * nearer pixel + 1/4 * further pixel in each */
        /* dimension, thus 9/16, 3/16, 3/16, 1/16 overall */
-        nextcolsum = GETJSAMPLE(*inptr0++) * 3 + GETJSAMPLE(*inptr1++);
+        nextcolsum = (*inptr0++) * 3 + (*inptr1++);
        *outptr++ = (JSAMPLE)((thiscolsum * 3 + lastcolsum + 8) >> 4);
        *outptr++ = (JSAMPLE)((thiscolsum * 3 + nextcolsum + 7) >> 4);
        lastcolsum = thiscolsum;  thiscolsum = nextcolsum;
@@ -477,7 +477,13 @@ jinit_upsampler(j_decompress_ptr cinfo)
    } else if (h_in_group == h_out_group &&
               v_in_group * 2 == v_out_group && do_fancy) {
      /* Non-fancy upsampling is handled by the generic method */
-      upsample->methods[ci] = h1v2_fancy_upsample;
+#if defined(__arm__) || defined(__aarch64__) || \
+    defined(_M_ARM) || defined(_M_ARM64)
+      if (jsimd_can_h1v2_fancy_upsample())
+        upsample->methods[ci] = jsimd_h1v2_fancy_upsample;
+      else
+#endif
+        upsample->methods[ci] = h1v2_fancy_upsample;
      upsample->pub.need_context_rows = TRUE;
    } else if (h_in_group * 2 == h_out_group &&
               v_in_group * 2 == v_out_group) {
@@ -207,6 +207,10 @@ JMESSAGE(JWRN_ARITH_BAD_CODE, "Corrupt JPEG data: bad arithmetic code")
 #endif
 #endif
 JMESSAGE(JWRN_BOGUS_ICC, "Corrupt JPEG data: bad ICC marker")
+#if JPEG_LIB_VERSION < 70
+JMESSAGE(JERR_BAD_DROP_SAMPLING,
+         "Component index %d: mismatching sampling ratio %d:%d, %d:%d, %c")
+#endif

 #ifdef JMAKE_ENUM_LIST

@@ -252,6 +256,15 @@ JMESSAGE(JWRN_BOGUS_ICC, "Corrupt JPEG data: bad ICC marker")
   (cinfo)->err->msg_parm.i[2] = (p3), \
   (cinfo)->err->msg_parm.i[3] = (p4), \
   (*(cinfo)->err->error_exit) ((j_common_ptr)(cinfo)))
+#define ERREXIT6(cinfo, code, p1, p2, p3, p4, p5, p6) \
+  ((cinfo)->err->msg_code = (code), \
+   (cinfo)->err->msg_parm.i[0] = (p1), \
+   (cinfo)->err->msg_parm.i[1] = (p2), \
+   (cinfo)->err->msg_parm.i[2] = (p3), \
+   (cinfo)->err->msg_parm.i[3] = (p4), \
+   (cinfo)->err->msg_parm.i[4] = (p5), \
+   (cinfo)->err->msg_parm.i[5] = (p6), \
+   (*(cinfo)->err->error_exit) ((j_common_ptr)(cinfo)))
 #define ERREXITS(cinfo, code, str) \
  ((cinfo)->err->msg_code = (code), \
   strncpy((cinfo)->err->msg_parm.s, (str), JMSG_STR_PARM_MAX), \
@@ -3,7 +3,7 @@
 *
 * This file was part of the Independent JPEG Group's software:
 * Copyright (C) 1991-1998, Thomas G. Lane.
- * Modification developed 2002-2009 by Guido Vollbeding.
+ * Modification developed 2002-2018 by Guido Vollbeding.
 * libjpeg-turbo Modifications:
 * Copyright (C) 2015, 2020, D. R. Commander.
 * For conditions of distribution and use, see the accompanying README.ijg
@@ -417,7 +417,7 @@ jpeg_idct_islow(j_decompress_ptr cinfo, jpeg_component_info *compptr,

 /*
 * Perform dequantization and inverse DCT on one block of coefficients,
- * producing a 7x7 output block.
+ * producing a reduced-size 7x7 output block.
 *
 * Optimized algorithm with 12 multiplications in the 1-D kernel.
 * cK represents sqrt(2) * cos(K*pi/14).
@@ -1258,7 +1258,7 @@ jpeg_idct_10x10(j_decompress_ptr cinfo, jpeg_component_info *compptr,

 /*
 * Perform dequantization and inverse DCT on one block of coefficients,
- * producing a 11x11 output block.
+ * producing an 11x11 output block.
 *
 * Optimized algorithm with 24 multiplications in the 1-D kernel.
 * cK represents sqrt(2) * cos(K*pi/22).
@@ -2398,7 +2398,7 @@ jpeg_idct_16x16(j_decompress_ptr cinfo, jpeg_component_info *compptr,
    tmp0 = DEQUANTIZE(inptr[DCTSIZE * 0], quantptr[DCTSIZE * 0]);
    tmp0 = LEFT_SHIFT(tmp0, CONST_BITS);
    /* Add fudge factor here for final descale. */
-    tmp0 += 1 << (CONST_BITS - PASS1_BITS - 1);
+    tmp0 += ONE << (CONST_BITS - PASS1_BITS - 1);

    z1 = DEQUANTIZE(inptr[DCTSIZE * 4], quantptr[DCTSIZE * 4]);
    tmp1 = MULTIPLY(z1, FIX(1.306562965));      /* c4[16] = c2[8] */
@@ -4,7 +4,7 @@
 * This file was part of the Independent JPEG Group's software:
 * Copyright (C) 1991-1997, Thomas G. Lane.
 * libjpeg-turbo Modifications:
- * Copyright (C) 2016, D. R. Commander.
+ * Copyright (C) 2016, 2021, D. R. Commander.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -1032,7 +1032,7 @@ free_pool(j_common_ptr cinfo, int pool_id)
    large_pool_ptr next_lhdr_ptr = lhdr_ptr->next;
    space_freed = lhdr_ptr->bytes_used +
                  lhdr_ptr->bytes_left +
-                  sizeof(large_pool_hdr);
+                  sizeof(large_pool_hdr) + ALIGN_SIZE - 1;
    jpeg_free_large(cinfo, (void *)lhdr_ptr, space_freed);
    mem->total_space_allocated -= space_freed;
    lhdr_ptr = next_lhdr_ptr;
@@ -1045,7 +1045,7 @@ free_pool(j_common_ptr cinfo, int pool_id)
  while (shdr_ptr != NULL) {
    small_pool_ptr next_shdr_ptr = shdr_ptr->next;
    space_freed = shdr_ptr->bytes_used + shdr_ptr->bytes_left +
-                  sizeof(small_pool_hdr);
+                  sizeof(small_pool_hdr) + ALIGN_SIZE - 1;
    jpeg_free_small(cinfo, (void *)shdr_ptr, space_freed);
    mem->total_space_allocated -= space_freed;
    shdr_ptr = next_shdr_ptr;
@@ -43,25 +43,11 @@

 #if BITS_IN_JSAMPLE == 8
 /* JSAMPLE should be the smallest type that will hold the values 0..255.
- * You can use a signed char by having GETJSAMPLE mask it with 0xFF.
 */

-#ifdef HAVE_UNSIGNED_CHAR
-
 typedef unsigned char JSAMPLE;
 #define GETJSAMPLE(value)  ((int)(value))

-#else /* not HAVE_UNSIGNED_CHAR */
-
-typedef char JSAMPLE;
-#ifdef __CHAR_UNSIGNED__
-#define GETJSAMPLE(value)  ((int)(value))
-#else
-#define GETJSAMPLE(value)  ((int)(value) & 0xFF)
-#endif /* __CHAR_UNSIGNED__ */
-
-#endif /* HAVE_UNSIGNED_CHAR */
-
 #define MAXJSAMPLE      255
 #define CENTERJSAMPLE   128

@@ -97,22 +83,9 @@ typedef short JCOEF;
 * managers, this is also the data type passed to fread/fwrite.
 */

-#ifdef HAVE_UNSIGNED_CHAR
-
 typedef unsigned char JOCTET;
 #define GETJOCTET(value)  (value)

-#else /* not HAVE_UNSIGNED_CHAR */
-
-typedef char JOCTET;
-#ifdef __CHAR_UNSIGNED__
-#define GETJOCTET(value)  (value)
-#else
-#define GETJOCTET(value)  ((value) & 0xFF)
-#endif /* __CHAR_UNSIGNED__ */
-
-#endif /* HAVE_UNSIGNED_CHAR */
-

 /* These typedefs are used for various table entries and so forth.
 * They must be at least as wide as specified; but making them too big
@@ -123,15 +96,7 @@ typedef char JOCTET;

 /* UINT8 must hold at least the values 0..255. */

-#ifdef HAVE_UNSIGNED_CHAR
 typedef unsigned char UINT8;
-#else /* not HAVE_UNSIGNED_CHAR */
-#ifdef __CHAR_UNSIGNED__
-typedef char UINT8;
-#else /* not __CHAR_UNSIGNED__ */
-typedef short UINT8;
-#endif /* __CHAR_UNSIGNED__ */
-#endif /* HAVE_UNSIGNED_CHAR */

 /* UINT16 must hold at least the values 0..65535. */

@@ -5,8 +5,9 @@
 * Copyright (C) 1991-1997, Thomas G. Lane.
 * Modified 1997-2009 by Guido Vollbeding.
 * libjpeg-turbo Modifications:
- * Copyright (C) 2015-2016, D. R. Commander.
+ * Copyright (C) 2015-2016, 2019, 2021, D. R. Commander.
 * Copyright (C) 2015, Google, Inc.
+ * Copyright (C) 2021, Alex Richardson.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -47,6 +48,18 @@ typedef enum {            /* Operating modes for buffer controllers */
 /* JLONG must hold at least signed 32-bit values. */
 typedef long JLONG;

+/* JUINTPTR must hold pointer values. */
+#ifdef __UINTPTR_TYPE__
+/*
+ * __UINTPTR_TYPE__ is GNU-specific and available in GCC 4.6+ and Clang 3.0+.
+ * Fortunately, that is sufficient to support the few architectures for which
+ * sizeof(void *) != sizeof(size_t).  The only other options would require C99
+ * or Clang-specific builtins.
+ */
+typedef __UINTPTR_TYPE__ JUINTPTR;
+#else
+typedef size_t JUINTPTR;
+#endif

 /*
 * Left shift macro that handles a negative operand without causing any
@@ -158,6 +171,9 @@ struct jpeg_decomp_master {
  JDIMENSION first_MCU_col[MAX_COMPONENTS];
  JDIMENSION last_MCU_col[MAX_COMPONENTS];
  boolean jinit_upsampler_no_alloc;
+
+  /* Last iMCU row that was successfully decoded */
+  JDIMENSION last_good_iMCU_row;
 };

 /* Input control module */
@@ -479,7 +479,7 @@ color_quantize(j_decompress_ptr cinfo, JSAMPARRAY input_buf,
    for (col = width; col > 0; col--) {
      pixcode = 0;
      for (ci = 0; ci < nc; ci++) {
-        pixcode += GETJSAMPLE(colorindex[ci][GETJSAMPLE(*ptrin++)]);
+        pixcode += colorindex[ci][*ptrin++];
      }
      *ptrout++ = (JSAMPLE)pixcode;
    }
@@ -506,9 +506,9 @@ color_quantize3(j_decompress_ptr cinfo, JSAMPARRAY input_buf,
    ptrin = input_buf[row];
    ptrout = output_buf[row];
    for (col = width; col > 0; col--) {
-      pixcode  = GETJSAMPLE(colorindex0[GETJSAMPLE(*ptrin++)]);
-      pixcode += GETJSAMPLE(colorindex1[GETJSAMPLE(*ptrin++)]);
-      pixcode += GETJSAMPLE(colorindex2[GETJSAMPLE(*ptrin++)]);
+      pixcode  = colorindex0[*ptrin++];
+      pixcode += colorindex1[*ptrin++];
+      pixcode += colorindex2[*ptrin++];
      *ptrout++ = (JSAMPLE)pixcode;
    }
  }
@@ -552,7 +552,7 @@ quantize_ord_dither(j_decompress_ptr cinfo, JSAMPARRAY input_buf,
         * required amount of padding.
         */
        *output_ptr +=
-          colorindex_ci[GETJSAMPLE(*input_ptr) + dither[col_index]];
+          colorindex_ci[*input_ptr + dither[col_index]];
        input_ptr += nc;
        output_ptr++;
        col_index = (col_index + 1) & ODITHER_MASK;
@@ -595,12 +595,9 @@ quantize3_ord_dither(j_decompress_ptr cinfo, JSAMPARRAY input_buf,
    col_index = 0;

    for (col = width; col > 0; col--) {
-      pixcode  =
-        GETJSAMPLE(colorindex0[GETJSAMPLE(*input_ptr++) + dither0[col_index]]);
-      pixcode +=
-        GETJSAMPLE(colorindex1[GETJSAMPLE(*input_ptr++) + dither1[col_index]]);
-      pixcode +=
-        GETJSAMPLE(colorindex2[GETJSAMPLE(*input_ptr++) + dither2[col_index]]);
+      pixcode  = colorindex0[(*input_ptr++) + dither0[col_index]];
+      pixcode += colorindex1[(*input_ptr++) + dither1[col_index]];
+      pixcode += colorindex2[(*input_ptr++) + dither2[col_index]];
      *output_ptr++ = (JSAMPLE)pixcode;
      col_index = (col_index + 1) & ODITHER_MASK;
    }
@@ -677,15 +674,15 @@ quantize_fs_dither(j_decompress_ptr cinfo, JSAMPARRAY input_buf,
         * The maximum error is +- MAXJSAMPLE; this sets the required size
         * of the range_limit array.
         */
-        cur += GETJSAMPLE(*input_ptr);
-        cur = GETJSAMPLE(range_limit[cur]);
+        cur += *input_ptr;
+        cur = range_limit[cur];
        /* Select output value, accumulate into output code for this pixel */
-        pixcode = GETJSAMPLE(colorindex_ci[cur]);
+        pixcode = colorindex_ci[cur];
        *output_ptr += (JSAMPLE)pixcode;
        /* Compute actual representation error at this pixel */
        /* Note: we can do this even though we don't have the final */
        /* pixel code, because the colormap is orthogonal. */
-        cur -= GETJSAMPLE(colormap_ci[pixcode]);
+        cur -= colormap_ci[pixcode];
        /* Compute error fractions to be propagated to adjacent pixels.
         * Add these into the running sums, and simultaneously shift the
         * next-line error sums left by 1 column.
@@ -215,9 +215,9 @@ prescan_quantize(j_decompress_ptr cinfo, JSAMPARRAY input_buf,
    ptr = input_buf[row];
    for (col = width; col > 0; col--) {
      /* get pixel value and index into the histogram */
-      histp = &histogram[GETJSAMPLE(ptr[0]) >> C0_SHIFT]
-                        [GETJSAMPLE(ptr[1]) >> C1_SHIFT]
-                        [GETJSAMPLE(ptr[2]) >> C2_SHIFT];
+      histp = &histogram[ptr[0] >> C0_SHIFT]
+                        [ptr[1] >> C1_SHIFT]
+                        [ptr[2] >> C2_SHIFT];
      /* increment, check for overflow and undo increment if so. */
      if (++(*histp) <= 0)
        (*histp)--;
@@ -665,7 +665,7 @@ find_nearby_colors(j_decompress_ptr cinfo, int minc0, int minc1, int minc2,

  for (i = 0; i < numcolors; i++) {
    /* We compute the squared-c0-distance term, then add in the other two. */
-    x = GETJSAMPLE(cinfo->colormap[0][i]);
+    x = cinfo->colormap[0][i];
    if (x < minc0) {
      tdist = (x - minc0) * C0_SCALE;
      min_dist = tdist * tdist;
@@ -688,7 +688,7 @@ find_nearby_colors(j_decompress_ptr cinfo, int minc0, int minc1, int minc2,
      }
    }

-    x = GETJSAMPLE(cinfo->colormap[1][i]);
+    x = cinfo->colormap[1][i];
    if (x < minc1) {
      tdist = (x - minc1) * C1_SCALE;
      min_dist += tdist * tdist;
@@ -710,7 +710,7 @@ find_nearby_colors(j_decompress_ptr cinfo, int minc0, int minc1, int minc2,
      }
    }

-    x = GETJSAMPLE(cinfo->colormap[2][i]);
+    x = cinfo->colormap[2][i];
    if (x < minc2) {
      tdist = (x - minc2) * C2_SCALE;
      min_dist += tdist * tdist;
@@ -788,13 +788,13 @@ find_best_colors(j_decompress_ptr cinfo, int minc0, int minc1, int minc2,
 #define STEP_C2  ((1 << C2_SHIFT) * C2_SCALE)

  for (i = 0; i < numcolors; i++) {
-    icolor = GETJSAMPLE(colorlist[i]);
+    icolor = colorlist[i];
    /* Compute (square of) distance from minc0/c1/c2 to this color */
-    inc0 = (minc0 - GETJSAMPLE(cinfo->colormap[0][icolor])) * C0_SCALE;
+    inc0 = (minc0 - cinfo->colormap[0][icolor]) * C0_SCALE;
    dist0 = inc0 * inc0;
-    inc1 = (minc1 - GETJSAMPLE(cinfo->colormap[1][icolor])) * C1_SCALE;
+    inc1 = (minc1 - cinfo->colormap[1][icolor]) * C1_SCALE;
    dist0 += inc1 * inc1;
-    inc2 = (minc2 - GETJSAMPLE(cinfo->colormap[2][icolor])) * C2_SCALE;
+    inc2 = (minc2 - cinfo->colormap[2][icolor]) * C2_SCALE;
    dist0 += inc2 * inc2;
    /* Form the initial difference increments */
    inc0 = inc0 * (2 * STEP_C0) + STEP_C0 * STEP_C0;
@@ -879,7 +879,7 @@ fill_inverse_cmap(j_decompress_ptr cinfo, int c0, int c1, int c2)
    for (ic1 = 0; ic1 < BOX_C1_ELEMS; ic1++) {
      cachep = &histogram[c0 + ic0][c1 + ic1][c2];
      for (ic2 = 0; ic2 < BOX_C2_ELEMS; ic2++) {
-        *cachep++ = (histcell)(GETJSAMPLE(*cptr++) + 1);
+        *cachep++ = (histcell)((*cptr++) + 1);
      }
    }
  }
@@ -909,9 +909,9 @@ pass2_no_dither(j_decompress_ptr cinfo, JSAMPARRAY input_buf,
    outptr = output_buf[row];
    for (col = width; col > 0; col--) {
      /* get pixel value and index into the cache */
-      c0 = GETJSAMPLE(*inptr++) >> C0_SHIFT;
-      c1 = GETJSAMPLE(*inptr++) >> C1_SHIFT;
-      c2 = GETJSAMPLE(*inptr++) >> C2_SHIFT;
+      c0 = (*inptr++) >> C0_SHIFT;
+      c1 = (*inptr++) >> C1_SHIFT;
+      c2 = (*inptr++) >> C2_SHIFT;
      cachep = &histogram[c0][c1][c2];
      /* If we have not seen this color before, find nearest colormap entry */
      /* and update the cache */
@@ -996,12 +996,12 @@ pass2_fs_dither(j_decompress_ptr cinfo, JSAMPARRAY input_buf,
       * The maximum error is +- MAXJSAMPLE (or less with error limiting);
       * this sets the required size of the range_limit array.
       */
-      cur0 += GETJSAMPLE(inptr[0]);
-      cur1 += GETJSAMPLE(inptr[1]);
-      cur2 += GETJSAMPLE(inptr[2]);
-      cur0 = GETJSAMPLE(range_limit[cur0]);
-      cur1 = GETJSAMPLE(range_limit[cur1]);
-      cur2 = GETJSAMPLE(range_limit[cur2]);
+      cur0 += inptr[0];
+      cur1 += inptr[1];
+      cur2 += inptr[2];
+      cur0 = range_limit[cur0];
+      cur1 = range_limit[cur1];
+      cur2 = range_limit[cur2];
      /* Index into the cache with adjusted pixel value */
      cachep =
        &histogram[cur0 >> C0_SHIFT][cur1 >> C1_SHIFT][cur2 >> C2_SHIFT];
@@ -1015,9 +1015,9 @@ pass2_fs_dither(j_decompress_ptr cinfo, JSAMPARRAY input_buf,
        register int pixcode = *cachep - 1;
        *outptr = (JSAMPLE)pixcode;
        /* Compute representation error for this pixel */
-        cur0 -= GETJSAMPLE(colormap0[pixcode]);
-        cur1 -= GETJSAMPLE(colormap1[pixcode]);
-        cur2 -= GETJSAMPLE(colormap2[pixcode]);
+        cur0 -= colormap0[pixcode];
+        cur1 -= colormap1[pixcode];
+        cur2 -= colormap2[pixcode];
      }
      /* Compute error fractions to be propagated to adjacent pixels.
       * Add these into the running sums, and simultaneously shift the
@@ -4,6 +4,7 @@
 * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
 * Copyright (C) 2011, 2014, D. R. Commander.
 * Copyright (C) 2015-2016, 2018, Matthieu Darbois.
+ * Copyright (C) 2020, Arm Limited.
 *
 * Based on the x86 SIMD extension for IJG JPEG library,
 * Copyright (C) 1999-2006, MIYASAKA Masaru.
@@ -75,6 +76,7 @@ EXTERN(void) jsimd_int_upsample(j_decompress_ptr cinfo,

 EXTERN(int) jsimd_can_h2v2_fancy_upsample(void);
 EXTERN(int) jsimd_can_h2v1_fancy_upsample(void);
+EXTERN(int) jsimd_can_h1v2_fancy_upsample(void);

 EXTERN(void) jsimd_h2v2_fancy_upsample(j_decompress_ptr cinfo,
                                       jpeg_component_info *compptr,
@@ -84,6 +86,10 @@ EXTERN(void) jsimd_h2v1_fancy_upsample(j_decompress_ptr cinfo,
                                       jpeg_component_info *compptr,
                                       JSAMPARRAY input_data,
                                       JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h1v2_fancy_upsample(j_decompress_ptr cinfo,
+                                       jpeg_component_info *compptr,
+                                       JSAMPARRAY input_data,
+                                       JSAMPARRAY *output_data_ptr);

 EXTERN(int) jsimd_can_h2v2_merged_upsample(void);
 EXTERN(int) jsimd_can_h2v1_merged_upsample(void);
@@ -4,6 +4,7 @@
 * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
 * Copyright (C) 2009-2011, 2014, D. R. Commander.
 * Copyright (C) 2015-2016, 2018, Matthieu Darbois.
+ * Copyright (C) 2020, Arm Limited.
 *
 * Based on the x86 SIMD extension for IJG JPEG library,
 * Copyright (C) 1999-2006, MIYASAKA Masaru.
@@ -169,6 +170,12 @@ jsimd_can_h2v1_fancy_upsample(void)
  return 0;
 }

+GLOBAL(int)
+jsimd_can_h1v2_fancy_upsample(void)
+{
+  return 0;
+}
+
 GLOBAL(void)
 jsimd_h2v2_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
                          JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
@@ -181,6 +188,12 @@ jsimd_h2v1_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
 {
 }

+GLOBAL(void)
+jsimd_h1v2_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                          JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+}
+
 GLOBAL(int)
 jsimd_can_h2v2_merged_upsample(void)
 {
@@ -2,9 +2,9 @@
 * jversion.h
 *
 * This file was part of the Independent JPEG Group's software:
- * Copyright (C) 1991-2012, Thomas G. Lane, Guido Vollbeding.
+ * Copyright (C) 1991-2020, Thomas G. Lane, Guido Vollbeding.
 * libjpeg-turbo Modifications:
- * Copyright (C) 2010, 2012-2020, D. R. Commander.
+ * Copyright (C) 2010, 2012-2021, D. R. Commander.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -37,9 +37,9 @@
 */

 #define JCOPYRIGHT \
-  "Copyright (C) 2009-2020 D. R. Commander\n" \
+  "Copyright (C) 2009-2021 D. R. Commander\n" \
  "Copyright (C) 2015, 2020 Google, Inc.\n" \
-  "Copyright (C) 2019 Arm Limited\n" \
+  "Copyright (C) 2019-2020 Arm Limited\n" \
  "Copyright (C) 2015-2016, 2018 Matthieu Darbois\n" \
  "Copyright (C) 2011-2016 Siarhei Siamashka\n" \
  "Copyright (C) 2015 Intel Corporation\n" \
@@ -48,7 +48,7 @@
  "Copyright (C) 2009, 2012 Pierre Ossman for Cendio AB\n" \
  "Copyright (C) 2009-2011 Nokia Corporation and/or its subsidiary(-ies)\n" \
  "Copyright (C) 1999-2006 MIYASAKA Masaru\n" \
-  "Copyright (C) 1991-2017 Thomas G. Lane, Guido Vollbeding"
+  "Copyright (C) 1991-2020 Thomas G. Lane, Guido Vollbeding"

 #define JCOPYRIGHT_SHORT \
-  "Copyright (C) 1991-2020 The libjpeg-turbo Project and many others"
+  "Copyright (C) 1991-2021 The libjpeg-turbo Project and many others"
@@ -239,6 +239,9 @@ if(HOST_BIG_ENDIAN)
 else()
  set(HOST_BIG_ENDIAN 0)
 endif()
+if(HOST_BIG_ENDIAN)
+  add_definitions(-DWORDS_BIGENDIAN)
+endif()

 # IEEE floating point
 set(HAVE_IEEEFP 1 CACHE STRING "IEEE floating point is available")
@@ -24,6 +24,10 @@
 #ifndef _LIBPORT_
 #define	_LIBPORT_

+#if defined(HAVE_CONFIG_H)
+#  include <tif_config.h>
+#endif
+
 int getopt(int argc, char * const argv[], const char *optstring);
 extern   char *optarg;
 extern   int opterr;
@@ -36,16 +40,16 @@ int strcasecmp(const char *s1, const char *s2);
 #  define HAVE_GETOPT 1
 #endif

-#if HAVE_STRTOL
+#if !defined(HAVE_STRTOL)
 long strtol(const char *nptr, char **endptr, int base);
 #endif
-#if HAVE_STRTOLL
+#if !defined(HAVE_STRTOLL)
 long long strtoll(const char *nptr, char **endptr, int base);
 #endif
-#if HAVE_STRTOUL
+#if !defined(HAVE_STRTOUL)
 unsigned long strtoul(const char *nptr, char **endptr, int base);
 #endif
-#if HAVE_STRTOULL
+#if !defined(HAVE_STRTOULL)
 unsigned long long strtoull(const char *nptr, char **endptr, int base);
 #endif

@@ -35,27 +35,61 @@
 uint32
 _TIFFMultiply32(TIFF* tif, uint32 first, uint32 second, const char* where)
 {
-	uint32 bytes = first * second;
-
-	if (second && bytes / second != first) {
+	if (second && first > TIFF_UINT32_MAX / second) {
 		TIFFErrorExt(tif->tif_clientdata, where, "Integer overflow in %s", where);
-		bytes = 0;
+		return 0;
 	}

-	return bytes;
+	return first * second;
 }

 uint64
 _TIFFMultiply64(TIFF* tif, uint64 first, uint64 second, const char* where)
 {
-	uint64 bytes = first * second;
-
-	if (second && bytes / second != first) {
+	if (second && first > TIFF_UINT64_MAX / second) {
 		TIFFErrorExt(tif->tif_clientdata, where, "Integer overflow in %s", where);
-		bytes = 0;
+		return 0;
 	}

-	return bytes;
+	return first * second;
+}
+
+tmsize_t
+_TIFFMultiplySSize(TIFF* tif, tmsize_t first, tmsize_t second, const char* where)
+{
+    if( first <= 0 || second <= 0 )
+    {
+        if( tif != NULL && where != NULL )
+        {
+            TIFFErrorExt(tif->tif_clientdata, where,
+                        "Invalid argument to _TIFFMultiplySSize() in %s", where);
+        }
+        return 0;
+    }
+
+    if( first > TIFF_TMSIZE_T_MAX / second )
+    {
+        if( tif != NULL && where != NULL )
+        {
+            TIFFErrorExt(tif->tif_clientdata, where,
+                        "Integer overflow in %s", where);
+        }
+        return 0;
+    }
+    return first * second;
+}
+
+tmsize_t _TIFFCastUInt64ToSSize(TIFF* tif, uint64 val, const char* module)
+{
+    if( val > (uint64)TIFF_TMSIZE_T_MAX )
+    {
+        if( tif != NULL && module != NULL )
+        {
+            TIFFErrorExt(tif->tif_clientdata,module,"Integer overflow");
+        }
+        return 0;
+    }
+    return (tmsize_t)val;
 }

 void*
@@ -63,13 +97,14 @@ _TIFFCheckRealloc(TIFF* tif, void* buffer,
 		  tmsize_t nmemb, tmsize_t elem_size, const char* what)
 {
 	void* cp = NULL;
-	tmsize_t bytes = nmemb * elem_size;
-
+        tmsize_t count = _TIFFMultiplySSize(tif, nmemb, elem_size, NULL);
 	/*
-	 * XXX: Check for integer overflow.
+	 * Check for integer overflow.
 	 */
-	if (nmemb && elem_size && bytes / elem_size == nmemb)
-		cp = _TIFFrealloc(buffer, bytes);
+	if (count != 0)
+	{
+		cp = _TIFFrealloc(buffer, count);
+	}

 	if (cp == NULL) {
 		TIFFErrorExt(tif->tif_clientdata, tif->tif_name,
@@ -235,7 +270,7 @@ TIFFVGetFieldDefaulted(TIFF* tif, uint32 tag, va_list ap)
 		return (1);
 	case TIFFTAG_EXTRASAMPLES:
 		*va_arg(ap, uint16 *) = td->td_extrasamples;
-		*va_arg(ap, uint16 **) = td->td_sampleinfo;
+		*va_arg(ap, const uint16 **) = td->td_sampleinfo;
 		return (1);
 	case TIFFTAG_MATTEING:
 		*va_arg(ap, uint16 *) =
@@ -257,8 +292,8 @@ TIFFVGetFieldDefaulted(TIFF* tif, uint32 tag, va_list ap)
 	case TIFFTAG_YCBCRCOEFFICIENTS:
 		{
 			/* defaults are from CCIR Recommendation 601-1 */
-			static float ycbcrcoeffs[] = { 0.299f, 0.587f, 0.114f };
-			*va_arg(ap, float **) = ycbcrcoeffs;
+			static const float ycbcrcoeffs[] = { 0.299f, 0.587f, 0.114f };
+			*va_arg(ap, const float **) = ycbcrcoeffs;
 			return 1;
 		}
 	case TIFFTAG_YCBCRSUBSAMPLING:
@@ -270,14 +305,14 @@ TIFFVGetFieldDefaulted(TIFF* tif, uint32 tag, va_list ap)
 		return (1);
 	case TIFFTAG_WHITEPOINT:
 		{
-			static float whitepoint[2];
-
 			/* TIFF 6.0 specification tells that it is no default
 			   value for the WhitePoint, but AdobePhotoshop TIFF
 			   Technical Note tells that it should be CIE D50. */
-			whitepoint[0] =	D50_X0 / (D50_X0 + D50_Y0 + D50_Z0);
-			whitepoint[1] =	D50_Y0 / (D50_X0 + D50_Y0 + D50_Z0);
-			*va_arg(ap, float **) = whitepoint;
+			static const float whitepoint[] = {
+						D50_X0 / (D50_X0 + D50_Y0 + D50_Z0),
+						D50_Y0 / (D50_X0 + D50_Y0 + D50_Z0)
+			};
+			*va_arg(ap, const float **) = whitepoint;
 			return 1;
 		}
 	case TIFFTAG_TRANSFERFUNCTION:
@@ -286,16 +321,16 @@ TIFFVGetFieldDefaulted(TIFF* tif, uint32 tag, va_list ap)
 			TIFFErrorExt(tif->tif_clientdata, tif->tif_name, "No space for \"TransferFunction\" tag");
 			return (0);
 		}
-		*va_arg(ap, uint16 **) = td->td_transferfunction[0];
+		*va_arg(ap, const uint16 **) = td->td_transferfunction[0];
 		if (td->td_samplesperpixel - td->td_extrasamples > 1) {
-			*va_arg(ap, uint16 **) = td->td_transferfunction[1];
-			*va_arg(ap, uint16 **) = td->td_transferfunction[2];
+			*va_arg(ap, const uint16 **) = td->td_transferfunction[1];
+			*va_arg(ap, const uint16 **) = td->td_transferfunction[2];
 		}
 		return (1);
 	case TIFFTAG_REFERENCEBLACKWHITE:
 		if (!td->td_refblackwhite && !TIFFDefaultRefBlackWhite(td))
 			return (0);
-		*va_arg(ap, float **) = td->td_refblackwhite;
+		*va_arg(ap, const float **) = td->td_refblackwhite;
 		return (1);
 	}
 	return 0;
@@ -264,7 +264,7 @@ TIFFGetConfiguredCODECs()
 			return NULL;
 		}
 		codecs = new_codecs;
-		_TIFFmemcpy(codecs + i - 1, cd, sizeof(TIFFCodec));
+		_TIFFmemcpy(codecs + i - 1, cd->info, sizeof(TIFFCodec));
 		i++;
 	}
 	for (c = _TIFFBuiltinCODECS; c->name; c++) {
@@ -29,6 +29,7 @@
 * (and also some miscellaneous stuff)
 */
 #include "tiffiop.h"
+#include <float.h>	/*--: for Rational2Double */

 /*
 * These are used in the backwards compatibility code...
@@ -46,8 +47,8 @@ setByteArray(void** vpp, void* vp, size_t nmemb, size_t elem_size)
 		*vpp = 0;
 	}
 	if (vp) {
-		tmsize_t bytes = (tmsize_t)(nmemb * elem_size);
-		if (elem_size && bytes / elem_size == nmemb)
+		tmsize_t bytes = _TIFFMultiplySSize(NULL, nmemb, elem_size, NULL);
+		if (bytes)
 			*vpp = (void*) _TIFFmalloc(bytes);
 		if (*vpp)
 			_TIFFmemcpy(*vpp, vp, bytes);
@@ -123,7 +124,7 @@ setExtraSamples(TIFF* tif, va_list ap, uint32* v)
        {
                TIFFWarningExt(tif->tif_clientdata,module,
                    "ExtraSamples tag value is changing, "
-                    "but TransferFunction was read with a different value. Cancelling it");
+                    "but TransferFunction was read with a different value. Canceling it");
                TIFFClrFieldBit(tif,FIELD_TRANSFERFUNCTION);
                _TIFFfree(td->td_transferfunction[0]);
                td->td_transferfunction[0] = NULL;
@@ -205,7 +206,7 @@ _TIFFVSetField(TIFF* tif, uint32 tag, va_list ap)
 		/*
 		 * If the data require post-decoding processing to byte-swap
 		 * samples, set it up here.  Note that since tags are required
-		 * to be ordered, compression code can override this behaviour
+		 * to be ordered, compression code can override this behavior
 		 * in the setup method if it wants to roll the post decoding
 		 * work in with its normal work.
 		 */
@@ -275,7 +276,7 @@ _TIFFVSetField(TIFF* tif, uint32 tag, va_list ap)
            {
                TIFFWarningExt(tif->tif_clientdata,module,
                    "SamplesPerPixel tag value is changing, "
-                    "but SMinSampleValue tag was read with a different value. Cancelling it");
+                    "but SMinSampleValue tag was read with a different value. Canceling it");
                TIFFClrFieldBit(tif,FIELD_SMINSAMPLEVALUE);
                _TIFFfree(td->td_sminsamplevalue);
                td->td_sminsamplevalue = NULL;
@@ -284,7 +285,7 @@ _TIFFVSetField(TIFF* tif, uint32 tag, va_list ap)
            {
                TIFFWarningExt(tif->tif_clientdata,module,
                    "SamplesPerPixel tag value is changing, "
-                    "but SMaxSampleValue tag was read with a different value. Cancelling it");
+                    "but SMaxSampleValue tag was read with a different value. Canceling it");
                TIFFClrFieldBit(tif,FIELD_SMAXSAMPLEVALUE);
                _TIFFfree(td->td_smaxsamplevalue);
                td->td_smaxsamplevalue = NULL;
@@ -296,7 +297,7 @@ _TIFFVSetField(TIFF* tif, uint32 tag, va_list ap)
            {
                    TIFFWarningExt(tif->tif_clientdata,module,
                        "SamplesPerPixel tag value is changing, "
-                        "but TransferFunction was read with a different value. Cancelling it");
+                        "but TransferFunction was read with a different value. Canceling it");
                    TIFFClrFieldBit(tif,FIELD_TRANSFERFUNCTION);
                    _TIFFfree(td->td_transferfunction[0]);
                    td->td_transferfunction[0] = NULL;
@@ -393,7 +394,7 @@ _TIFFVSetField(TIFF* tif, uint32 tag, va_list ap)
 			if (tif->tif_mode != O_RDONLY)
 				goto badvalue32;
 			TIFFWarningExt(tif->tif_clientdata, tif->tif_name,
-				"Nonstandard tile width %d, convert file", v32);
+				"Nonstandard tile width %u, convert file", v32);
 		}
 		td->td_tilewidth = v32;
 		tif->tif_flags |= TIFF_ISTILED;
@@ -404,7 +405,7 @@ _TIFFVSetField(TIFF* tif, uint32 tag, va_list ap)
 			if (tif->tif_mode != O_RDONLY)
 				goto badvalue32;
 			TIFFWarningExt(tif->tif_clientdata, tif->tif_name,
-			    "Nonstandard tile length %d, convert file", v32);
+			    "Nonstandard tile length %u, convert file", v32);
 		}
 		td->td_tilelength = v32;
 		tif->tif_flags |= TIFF_ISTILED;
@@ -559,6 +560,10 @@ _TIFFVSetField(TIFF* tif, uint32 tag, va_list ap)
 		 * Set custom value ... save a copy of the custom tag value.
 		 */
 		tv_size = _TIFFDataSize(fip->field_type);
+		/*--: Rational2Double: For Rationals evaluate "set_field_type" to determine internal storage size. */
+		if (fip->field_type == TIFF_RATIONAL || fip->field_type == TIFF_SRATIONAL) {
+			tv_size = _TIFFSetGetFieldSize(fip->set_field_type);
+		}
 		if (tv_size == 0) {
 			status = 0;
 			TIFFErrorExt(tif->tif_clientdata, module,
@@ -638,6 +643,7 @@ _TIFFVSetField(TIFF* tif, uint32 tag, va_list ap)
 				  || fip->field_writecount == TIFF_VARIABLE2
 				  || fip->field_writecount == TIFF_SPP
 				  || tv->count > 1) {
+			  /*--: Rational2Double: For Rationals tv_size is set above to 4 or 8 according to fip->set_field_type! */
 				_TIFFmemcpy(tv->value, va_arg(ap, void *),
 				    tv->count * tv_size);
 			} else {
@@ -698,6 +704,22 @@ _TIFFVSetField(TIFF* tif, uint32 tag, va_list ap)
 					break;
 				case TIFF_RATIONAL:
 				case TIFF_SRATIONAL:
+					/*-- Rational2Double: For Rationals tv_size is set above to 4 or 8 according to fip->set_field_type! */
+					{
+						if (tv_size == 8) {
+							double v2 = va_arg(ap, double);
+							_TIFFmemcpy(val, &v2, tv_size);
+						} else {
+							/*-- default should be tv_size == 4 */
+							float v3 = (float)va_arg(ap, double);
+							_TIFFmemcpy(val, &v3, tv_size);
+							/*-- ToDo: After Testing, this should be removed and tv_size==4 should be set as default. */
+							if (tv_size != 4) {
+								TIFFErrorExt(0,"TIFFLib: _TIFFVSetField()", "Rational2Double: .set_field_type in not 4 but %d", tv_size); 
+							}
+						}
+					}
+					break;
 				case TIFF_FLOAT:
 					{
 						float v2 = _TIFFClampDoubleToFloat(va_arg(ap, double));
@@ -1011,19 +1033,19 @@ _TIFFVGetField(TIFF* tif, uint32 tag, va_list ap)
 			*va_arg(ap, uint16*) = td->td_halftonehints[1];
 			break;
 		case TIFFTAG_COLORMAP:
-			*va_arg(ap, uint16**) = td->td_colormap[0];
-			*va_arg(ap, uint16**) = td->td_colormap[1];
-			*va_arg(ap, uint16**) = td->td_colormap[2];
+			*va_arg(ap, const uint16**) = td->td_colormap[0];
+			*va_arg(ap, const uint16**) = td->td_colormap[1];
+			*va_arg(ap, const uint16**) = td->td_colormap[2];
 			break;
 		case TIFFTAG_STRIPOFFSETS:
 		case TIFFTAG_TILEOFFSETS:
 			_TIFFFillStriles( tif );
-			*va_arg(ap, uint64**) = td->td_stripoffset;
+			*va_arg(ap, const uint64**) = td->td_stripoffset_p;
 			break;
 		case TIFFTAG_STRIPBYTECOUNTS:
 		case TIFFTAG_TILEBYTECOUNTS:
 			_TIFFFillStriles( tif );
-			*va_arg(ap, uint64**) = td->td_stripbytecount;
+			*va_arg(ap, const uint64**) = td->td_stripbytecount_p;
 			break;
 		case TIFFTAG_MATTEING:
 			*va_arg(ap, uint16*) =
@@ -1032,7 +1054,7 @@ _TIFFVGetField(TIFF* tif, uint32 tag, va_list ap)
 			break;
 		case TIFFTAG_EXTRASAMPLES:
 			*va_arg(ap, uint16*) = td->td_extrasamples;
-			*va_arg(ap, uint16**) = td->td_sampleinfo;
+			*va_arg(ap, const uint16**) = td->td_sampleinfo;
 			break;
 		case TIFFTAG_TILEWIDTH:
 			*va_arg(ap, uint32*) = td->td_tilewidth;
@@ -1067,7 +1089,7 @@ _TIFFVGetField(TIFF* tif, uint32 tag, va_list ap)
 			break;
 		case TIFFTAG_SUBIFD:
 			*va_arg(ap, uint16*) = td->td_nsubifd;
-			*va_arg(ap, uint64**) = td->td_subifd;
+			*va_arg(ap, const uint64**) = td->td_subifd;
 			break;
 		case TIFFTAG_YCBCRPOSITIONING:
 			*va_arg(ap, uint16*) = td->td_ycbcrpositioning;
@@ -1077,20 +1099,20 @@ _TIFFVGetField(TIFF* tif, uint32 tag, va_list ap)
 			*va_arg(ap, uint16*) = td->td_ycbcrsubsampling[1];
 			break;
 		case TIFFTAG_TRANSFERFUNCTION:
-			*va_arg(ap, uint16**) = td->td_transferfunction[0];
+			*va_arg(ap, const uint16**) = td->td_transferfunction[0];
 			if (td->td_samplesperpixel - td->td_extrasamples > 1) {
-				*va_arg(ap, uint16**) = td->td_transferfunction[1];
-				*va_arg(ap, uint16**) = td->td_transferfunction[2];
+				*va_arg(ap, const uint16**) = td->td_transferfunction[1];
+				*va_arg(ap, const uint16**) = td->td_transferfunction[2];
 			} else {
-				*va_arg(ap, uint16**) = NULL;
-				*va_arg(ap, uint16**) = NULL;
+				*va_arg(ap, const uint16**) = NULL;
+				*va_arg(ap, const uint16**) = NULL;
 			}
 			break;
 		case TIFFTAG_REFERENCEBLACKWHITE:
-			*va_arg(ap, float**) = td->td_refblackwhite;
+			*va_arg(ap, const float**) = td->td_refblackwhite;
 			break;
 		case TIFFTAG_INKNAMES:
-			*va_arg(ap, char**) = td->td_inknames;
+			*va_arg(ap, const char**) = td->td_inknames;
 			break;
 		default:
 			{
@@ -1132,7 +1154,7 @@ _TIFFVGetField(TIFF* tif, uint32 tag, va_list ap)
 							*va_arg(ap, uint32*) = (uint32)tv->count;
 						else  /* Assume TIFF_VARIABLE */
 							*va_arg(ap, uint16*) = (uint16)tv->count;
-						*va_arg(ap, void **) = tv->value;
+						*va_arg(ap, const void **) = tv->value;
 						ret_val = 1;
 					} else if (fip->field_tag == TIFFTAG_DOTRANGE
 						   && strcmp(fip->field_name,"DotRange") == 0) {
@@ -1200,6 +1222,23 @@ _TIFFVGetField(TIFF* tif, uint32 tag, va_list ap)
 								break;
 							case TIFF_RATIONAL:
 							case TIFF_SRATIONAL:
+								{
+									/*-- Rational2Double: For Rationals evaluate "set_field_type" to determine internal storage size and return value size. */
+									int tv_size = _TIFFSetGetFieldSize(fip->set_field_type);
+									if (tv_size == 8) {
+										*va_arg(ap, double*) = *(double *)val;
+										ret_val = 1;
+									} else {
+										/*-- default should be tv_size == 4  */
+										*va_arg(ap, float*) = *(float *)val;
+										ret_val = 1;
+										/*-- ToDo: After Testing, this should be removed and tv_size==4 should be set as default. */
+										if (tv_size != 4) {
+											TIFFErrorExt(0,"TIFFLib: _TIFFVGetField()", "Rational2Double: .set_field_type in not 4 but %d", tv_size); 
+										}
+									}
+								}
+								break;
 							case TIFF_FLOAT:
 								*va_arg(ap, float*) =
 									*(float *)val;
@@ -1282,8 +1321,9 @@ TIFFFreeDirectory(TIFF* tif)
 	CleanupField(td_transferfunction[0]);
 	CleanupField(td_transferfunction[1]);
 	CleanupField(td_transferfunction[2]);
-	CleanupField(td_stripoffset);
-	CleanupField(td_stripbytecount);
+	CleanupField(td_stripoffset_p);
+	CleanupField(td_stripbytecount_p);
+        td->td_stripoffsetbyteallocsize = 0;
 	TIFFClrFieldBit(tif, FIELD_YCBCRSUBSAMPLING);
 	TIFFClrFieldBit(tif, FIELD_YCBCRPOSITIONING);

@@ -1296,10 +1336,8 @@ TIFFFreeDirectory(TIFF* tif)
 	td->td_customValueCount = 0;
 	CleanupField(td_customValues);

-#if defined(DEFER_STRILE_LOAD)
        _TIFFmemset( &(td->td_stripoffset_entry), 0, sizeof(TIFFDirEntry));
        _TIFFmemset( &(td->td_stripbytecount_entry), 0, sizeof(TIFFDirEntry));
-#endif        
 }
 #undef CleanupField

@@ -1365,6 +1403,17 @@ TIFFCreateEXIFDirectory(TIFF* tif)
 	return TIFFCreateCustomDirectory(tif, exifFieldArray);
 }

+/*
+ * Creates the EXIF GPS custom directory 
+ */
+int
+TIFFCreateGPSDirectory(TIFF* tif)
+{
+	const TIFFFieldArray* gpsFieldArray;
+	gpsFieldArray = _TIFFGetGpsFields();
+	return TIFFCreateCustomDirectory(tif, gpsFieldArray);
+}
+
 /*
 * Setup a default directory structure.
 */
@@ -1387,7 +1436,9 @@ TIFFDefaultDirectory(TIFF* tif)
 	td->td_tilewidth = 0;
 	td->td_tilelength = 0;
 	td->td_tiledepth = 1;
+#ifdef STRIPBYTECOUNTSORTED_UNUSED
 	td->td_stripbytecountsorted = 1; /* Our own arrays always sorted. */  
+#endif
 	td->td_resolutionunit = RESUNIT_INCH;
 	td->td_sampleformat = SAMPLEFORMAT_UINT;
 	td->td_imagedepth = 1;
@@ -58,6 +58,7 @@ typedef struct {
 		uint32 toff_long;
 		uint64 toff_long8;
 	} tdir_offset;		/* either offset or the data itself if fits */
+	uint8  tdir_ignore;	/* flag status to ignore tag when parsing tags in tif_dirread.c */
 } TIFFDirEntry;

 /*
@@ -97,13 +98,14 @@ typedef struct {
 	 * number of striles */
 	uint32  td_stripsperimage;  
 	uint32  td_nstrips;              /* size of offset & bytecount arrays */
-	uint64* td_stripoffset;
-	uint64* td_stripbytecount;
+	uint64* td_stripoffset_p;        /* should be accessed with TIFFGetStrileOffset */
+	uint64* td_stripbytecount_p;     /* should be accessed with TIFFGetStrileByteCount */
+        uint32  td_stripoffsetbyteallocsize; /* number of elements currently allocated for td_stripoffset/td_stripbytecount. Only used if TIFF_LAZYSTRILELOAD is set */
+#ifdef STRIPBYTECOUNTSORTED_UNUSED
 	int     td_stripbytecountsorted; /* is the bytecount array sorted ascending? */
-#if defined(DEFER_STRILE_LOAD)
+#endif
        TIFFDirEntry td_stripoffset_entry;    /* for deferred loading */
        TIFFDirEntry td_stripbytecount_entry; /* for deferred loading */
-#endif
 	uint16  td_nsubifd;
 	uint64* td_subifd;
 	/* YCbCr parameters */
@@ -118,6 +120,8 @@ typedef struct {

 	int     td_customValueCount;
        TIFFTagValue *td_customValues;
+
+        unsigned char td_deferstrilearraywriting; /* see TIFFDeferStrileArrayWriting() */
 } TIFFDirectory;

 /*
@@ -257,6 +261,7 @@ extern "C" {

 extern const TIFFFieldArray* _TIFFGetFields(void);
 extern const TIFFFieldArray* _TIFFGetExifFields(void);
+extern const TIFFFieldArray* _TIFFGetGpsFields(void);
 extern void _TIFFSetupFields(TIFF* tif, const TIFFFieldArray* infoarray);
 extern void _TIFFPrintFieldInfo(TIFF*, FILE*);

@@ -265,6 +270,7 @@ extern int _TIFFFillStriles(TIFF*);
 typedef enum {
 	tfiatImage,
 	tfiatExif,
+	tfiatGps,		/* EXIF-GPS fields array type */
 	tfiatOther
 } TIFFFieldArrayType;

@@ -47,9 +47,19 @@
 #endif
 static const TIFFFieldArray tiffFieldArray;
 static const TIFFFieldArray exifFieldArray;
+static const TIFFFieldArray gpsFieldArray;
 #ifdef _MSC_VER
 #pragma warning( pop )
 #endif
+/*--: Rational2Double: --
+ * The Rational2Double upgraded libtiff functionality allows the definition and achievement of true double-precision accuracy
+ * for TIFF tags of RATIONAL type and field_bit=FIELD_CUSTOM using the set_field_type = TIFF_SETGET_DOUBLE.
+ * Unfortunately, that changes the old implemented interface for TIFFGetField().
+ * In order to keep the old TIFFGetField() interface behavior those tags have to be redefined with set_field_type = TIFF_SETGET_FLOAT!
+ *
+ *  Rational custom arrays are already defined as _Cxx_FLOAT, thus can stay.
+ *
+ */

 static const TIFFField
 tiffFields[] = {
@@ -75,12 +85,12 @@ tiffFields[] = {
 	{ TIFFTAG_STRIPBYTECOUNTS, -1, -1, TIFF_LONG8, 0, TIFF_SETGET_UNDEFINED, TIFF_SETGET_UNDEFINED, FIELD_STRIPBYTECOUNTS, 0, 0, "StripByteCounts", NULL },
 	{ TIFFTAG_MINSAMPLEVALUE, -2, -1, TIFF_SHORT, 0, TIFF_SETGET_UINT16, TIFF_SETGET_UNDEFINED, FIELD_MINSAMPLEVALUE, 1, 0, "MinSampleValue", NULL },
 	{ TIFFTAG_MAXSAMPLEVALUE, -2, -1, TIFF_SHORT, 0, TIFF_SETGET_UINT16, TIFF_SETGET_UNDEFINED, FIELD_MAXSAMPLEVALUE, 1, 0, "MaxSampleValue", NULL },
-	{ TIFFTAG_XRESOLUTION, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_DOUBLE, TIFF_SETGET_UNDEFINED, FIELD_RESOLUTION, 1, 0, "XResolution", NULL },
-	{ TIFFTAG_YRESOLUTION, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_DOUBLE, TIFF_SETGET_UNDEFINED, FIELD_RESOLUTION, 1, 0, "YResolution", NULL },
+	{ TIFFTAG_XRESOLUTION, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_RESOLUTION, 1, 0, "XResolution", NULL },
+	{ TIFFTAG_YRESOLUTION, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_RESOLUTION, 1, 0, "YResolution", NULL },
 	{ TIFFTAG_PLANARCONFIG, 1, 1, TIFF_SHORT, 0, TIFF_SETGET_UINT16, TIFF_SETGET_UNDEFINED, FIELD_PLANARCONFIG, 0, 0, "PlanarConfiguration", NULL },
 	{ TIFFTAG_PAGENAME, -1, -1, TIFF_ASCII, 0, TIFF_SETGET_ASCII, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "PageName", NULL },
-	{ TIFFTAG_XPOSITION, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_DOUBLE, TIFF_SETGET_UNDEFINED, FIELD_POSITION, 1, 0, "XPosition", NULL },
-	{ TIFFTAG_YPOSITION, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_DOUBLE, TIFF_SETGET_UNDEFINED, FIELD_POSITION, 1, 0, "YPosition", NULL },
+	{ TIFFTAG_XPOSITION, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_POSITION, 1, 0, "XPosition", NULL },
+	{ TIFFTAG_YPOSITION, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_POSITION, 1, 0, "YPosition", NULL },
 	{ TIFFTAG_FREEOFFSETS, -1, -1, TIFF_LONG8, 0, TIFF_SETGET_UNDEFINED, TIFF_SETGET_UNDEFINED, FIELD_IGNORE, 0, 0, "FreeOffsets", NULL },
 	{ TIFFTAG_FREEBYTECOUNTS, -1, -1, TIFF_LONG8, 0, TIFF_SETGET_UNDEFINED, TIFF_SETGET_UNDEFINED, FIELD_IGNORE, 0, 0, "FreeByteCounts", NULL },
 	{ TIFFTAG_GRAYRESPONSEUNIT, 1, 1, TIFF_SHORT, 0, TIFF_SETGET_UNDEFINED, TIFF_SETGET_UNDEFINED, FIELD_IGNORE, 1, 0, "GrayResponseUnit", NULL },
@@ -135,14 +145,18 @@ tiffFields[] = {
 	{ TIFFTAG_PIXAR_MATRIX_WORLDTOSCREEN, 16, 16, TIFF_FLOAT, 0, TIFF_SETGET_C0_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "MatrixWorldToScreen", NULL },
 	{ TIFFTAG_PIXAR_MATRIX_WORLDTOCAMERA, 16, 16, TIFF_FLOAT, 0, TIFF_SETGET_C0_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "MatrixWorldToCamera", NULL },
 	{ TIFFTAG_CFAREPEATPATTERNDIM, 2, 2, TIFF_SHORT, 0, TIFF_SETGET_C0_UINT16, TIFF_SETGET_UNDEFINED,	FIELD_CUSTOM, 0,	0,	"CFARepeatPatternDim", NULL },
-	{ TIFFTAG_CFAPATTERN,	4, 4,	TIFF_BYTE, 0, TIFF_SETGET_C0_UINT8, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0,	0,	"CFAPattern" , NULL},
+	{ TIFFTAG_CFAPATTERN, -1, -1, TIFF_BYTE, 0, TIFF_SETGET_C16_UINT8, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 1, "CFAPattern" , NULL},
 	{ TIFFTAG_COPYRIGHT, -1, -1, TIFF_ASCII, 0, TIFF_SETGET_ASCII, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "Copyright", NULL },
 	/* end Pixar tags */
-	{ TIFFTAG_RICHTIFFIPTC, -3, -3, TIFF_LONG, 0, TIFF_SETGET_C32_UINT32, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 1, "RichTIFFIPTC", NULL },
+	{ TIFFTAG_RICHTIFFIPTC, -3, -3, TIFF_UNDEFINED, 0, TIFF_SETGET_C32_UINT8, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 1, "RichTIFFIPTC", NULL },
 	{ TIFFTAG_PHOTOSHOP, -3, -3, TIFF_BYTE, 0, TIFF_SETGET_C32_UINT8, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 1, "Photoshop", NULL },
-	{ TIFFTAG_EXIFIFD, 1, 1, TIFF_IFD8, 0, TIFF_SETGET_IFD8, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 0, "EXIFIFDOffset", (TIFFFieldArray*) &exifFieldArray },
+	/*--: EXIFIFD and GPSIFD specified as TIFF_LONG by Aware-Systems and not TIFF_IFD8 as in original LibTiff.
+	 *    However, for IFD-like tags, libtiff uses the data type TIFF_IFD8 in tiffFields[]-tag definition combined with
+	 *    a special handling procedure in order to write either a 32-bit value and the TIFF_IFD type-id into ClassicTIFF files 
+	 *    or a 64-bit value and the TIFF_IFD8 type-id into BigTIFF files. */
+	{ TIFFTAG_EXIFIFD, 1, 1, TIFF_IFD8, 0, TIFF_SETGET_IFD8, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "EXIFIFDOffset", (TIFFFieldArray*) &exifFieldArray },
 	{ TIFFTAG_ICCPROFILE, -3, -3, TIFF_UNDEFINED, 0, TIFF_SETGET_C32_UINT8, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 1, "ICC Profile", NULL },
-	{ TIFFTAG_GPSIFD, 1, 1, TIFF_IFD8, 0, TIFF_SETGET_IFD8, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 0, "GPSIFDOffset", NULL },
+	{ TIFFTAG_GPSIFD, 1, 1, TIFF_IFD8, 0, TIFF_SETGET_IFD8, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "GPSIFDOffset", (TIFFFieldArray*) &gpsFieldArray },
 	{ TIFFTAG_FAXRECVPARAMS, 1, 1, TIFF_LONG, 0, TIFF_SETGET_UINT32, TIFF_SETGET_UINT32, FIELD_CUSTOM, TRUE, FALSE, "FaxRecvParams", NULL },
 	{ TIFFTAG_FAXSUBADDRESS, -1, -1, TIFF_ASCII, 0, TIFF_SETGET_ASCII, TIFF_SETGET_ASCII, FIELD_CUSTOM, TRUE, FALSE, "FaxSubAddress", NULL },
 	{ TIFFTAG_FAXRECVTIME, 1, 1, TIFF_LONG, 0, TIFF_SETGET_UINT32, TIFF_SETGET_UINT32, FIELD_CUSTOM, TRUE, FALSE, "FaxRecvTime", NULL },
@@ -163,7 +177,7 @@ tiffFields[] = {
 	{ TIFFTAG_BLACKLEVELDELTAV, -1, -1, TIFF_SRATIONAL, 0, TIFF_SETGET_C16_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 1, "BlackLevelDeltaV", NULL },
 	{ TIFFTAG_WHITELEVEL, -1, -1, TIFF_LONG, 0, TIFF_SETGET_C16_UINT32, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 1, "WhiteLevel", NULL },
 	{ TIFFTAG_DEFAULTSCALE, 2, 2, TIFF_RATIONAL, 0, TIFF_SETGET_C0_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 0, "DefaultScale", NULL },
-	{ TIFFTAG_BESTQUALITYSCALE, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_DOUBLE, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 0, "BestQualityScale", NULL },
+	{ TIFFTAG_BESTQUALITYSCALE, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 0, "BestQualityScale", NULL },
 	{ TIFFTAG_DEFAULTCROPORIGIN, 2, 2, TIFF_RATIONAL, 0, TIFF_SETGET_C0_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 0, "DefaultCropOrigin", NULL },
 	{ TIFFTAG_DEFAULTCROPSIZE, 2, 2, TIFF_RATIONAL, 0, TIFF_SETGET_C0_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 0, "DefaultCropSize", NULL },
 	{ TIFFTAG_COLORMATRIX1, -1, -1, TIFF_SRATIONAL, 0, TIFF_SETGET_C16_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 1, "ColorMatrix1", NULL },
@@ -175,16 +189,16 @@ tiffFields[] = {
 	{ TIFFTAG_ANALOGBALANCE, -1, -1, TIFF_RATIONAL, 0, TIFF_SETGET_C16_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 1, "AnalogBalance", NULL },
 	{ TIFFTAG_ASSHOTNEUTRAL, -1, -1, TIFF_RATIONAL, 0, TIFF_SETGET_C16_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 1, "AsShotNeutral", NULL },
 	{ TIFFTAG_ASSHOTWHITEXY, 2, 2, TIFF_RATIONAL, 0, TIFF_SETGET_C0_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 0, "AsShotWhiteXY", NULL },
-	{ TIFFTAG_BASELINEEXPOSURE, 1, 1, TIFF_SRATIONAL, 0, TIFF_SETGET_DOUBLE, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 0, "BaselineExposure", NULL },
-	{ TIFFTAG_BASELINENOISE, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_DOUBLE, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 0, "BaselineNoise", NULL },
-	{ TIFFTAG_BASELINESHARPNESS, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_DOUBLE, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 0, "BaselineSharpness", NULL },
+	{ TIFFTAG_BASELINEEXPOSURE, 1, 1, TIFF_SRATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 0, "BaselineExposure", NULL },
+	{ TIFFTAG_BASELINENOISE, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 0, "BaselineNoise", NULL },
+	{ TIFFTAG_BASELINESHARPNESS, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 0, "BaselineSharpness", NULL },
 	{ TIFFTAG_BAYERGREENSPLIT, 1, 1, TIFF_LONG, 0, TIFF_SETGET_UINT32, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 0, "BayerGreenSplit", NULL },
-	{ TIFFTAG_LINEARRESPONSELIMIT, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_DOUBLE, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 0, "LinearResponseLimit", NULL },
+	{ TIFFTAG_LINEARRESPONSELIMIT, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 0, "LinearResponseLimit", NULL },
 	{ TIFFTAG_CAMERASERIALNUMBER, -1, -1, TIFF_ASCII, 0, TIFF_SETGET_ASCII, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "CameraSerialNumber", NULL },
 	{ TIFFTAG_LENSINFO, 4, 4, TIFF_RATIONAL, 0, TIFF_SETGET_C0_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 0, "LensInfo", NULL },
-	{ TIFFTAG_CHROMABLURRADIUS, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_DOUBLE, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 0, "ChromaBlurRadius", NULL },
-	{ TIFFTAG_ANTIALIASSTRENGTH, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_DOUBLE, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 0, "AntiAliasStrength", NULL },
-	{ TIFFTAG_SHADOWSCALE, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_DOUBLE, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 0, "ShadowScale", NULL },
+	{ TIFFTAG_CHROMABLURRADIUS, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 0, "ChromaBlurRadius", NULL },
+	{ TIFFTAG_ANTIALIASSTRENGTH, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 0, "AntiAliasStrength", NULL },
+	{ TIFFTAG_SHADOWSCALE, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 0, "ShadowScale", NULL },
 	{ TIFFTAG_DNGPRIVATEDATA, -1, -1, TIFF_BYTE, 0, TIFF_SETGET_C16_UINT8, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 1, "DNGPrivateData", NULL },
 	{ TIFFTAG_MAKERNOTESAFETY, 1, 1, TIFF_SHORT, 0, TIFF_SETGET_UINT16, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 0, "MakerNoteSafety", NULL },
 	{ TIFFTAG_CALIBRATIONILLUMINANT1, 1, 1, TIFF_SHORT, 0, TIFF_SETGET_UINT16, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 0, 0, "CalibrationIlluminant1", NULL },
@@ -217,47 +231,68 @@ tiffFields[] = {
 	/* begin pseudo tags */
 };

+/*
+ * EXIF tags  (Version 2.31, July 2016 plus version 2.32 May 2019)
+ */
 static const TIFFField
 exifFields[] = {
-	{ EXIFTAG_EXPOSURETIME, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_DOUBLE, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "ExposureTime", NULL },
-	{ EXIFTAG_FNUMBER, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_DOUBLE, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "FNumber", NULL },
+	{ EXIFTAG_EXPOSURETIME, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "ExposureTime", NULL },
+	{ EXIFTAG_FNUMBER, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "FNumber", NULL },
 	{ EXIFTAG_EXPOSUREPROGRAM, 1, 1, TIFF_SHORT, 0, TIFF_SETGET_UINT16, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "ExposureProgram", NULL },
 	{ EXIFTAG_SPECTRALSENSITIVITY, -1, -1, TIFF_ASCII, 0, TIFF_SETGET_ASCII, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "SpectralSensitivity", NULL },
 	{ EXIFTAG_ISOSPEEDRATINGS, -1, -1, TIFF_SHORT, 0, TIFF_SETGET_C16_UINT16, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 1, "ISOSpeedRatings", NULL },
 	{ EXIFTAG_OECF, -1, -1, TIFF_UNDEFINED, 0, TIFF_SETGET_C16_UINT8, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 1, "OptoelectricConversionFactor", NULL },
+	{ EXIFTAG_SENSITIVITYTYPE, 1, 1, TIFF_SHORT, 0, TIFF_SETGET_UINT16, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "SensitivityType", NULL },
+	{ EXIFTAG_STANDARDOUTPUTSENSITIVITY, 1, 1, TIFF_LONG, 0, TIFF_SETGET_UINT32, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "StandardOutputSensitivity", NULL },
+	{ EXIFTAG_RECOMMENDEDEXPOSUREINDEX, 1, 1, TIFF_LONG, 0, TIFF_SETGET_UINT32, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "RecommendedExposureIndex", NULL },
+	{ EXIFTAG_ISOSPEED, 1, 1, TIFF_LONG, 0, TIFF_SETGET_UINT32, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "ISOSpeed", NULL },
+	{ EXIFTAG_ISOSPEEDLATITUDEYYY, 1, 1, TIFF_LONG, 0, TIFF_SETGET_UINT32, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "ISOSpeedLatitudeyyy", NULL },
+	{ EXIFTAG_ISOSPEEDLATITUDEZZZ, 1, 1, TIFF_LONG, 0, TIFF_SETGET_UINT32, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "ISOSpeedLatitudezzz", NULL },
 	{ EXIFTAG_EXIFVERSION, 4, 4, TIFF_UNDEFINED, 0, TIFF_SETGET_C0_UINT8, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "ExifVersion", NULL },
 	{ EXIFTAG_DATETIMEORIGINAL, 20, 20, TIFF_ASCII, 0, TIFF_SETGET_ASCII, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "DateTimeOriginal", NULL },
 	{ EXIFTAG_DATETIMEDIGITIZED, 20, 20, TIFF_ASCII, 0, TIFF_SETGET_ASCII, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "DateTimeDigitized", NULL },
+	{ EXIFTAG_OFFSETTIME, 7, 7, TIFF_ASCII, 0, TIFF_SETGET_ASCII, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "OffsetTime", NULL },
+	{ EXIFTAG_OFFSETTIMEORIGINAL, 7, 7, TIFF_ASCII, 0, TIFF_SETGET_ASCII, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "OffsetTimeOriginal", NULL },
+	{ EXIFTAG_OFFSETTIMEDIGITIZED, 7, 7, TIFF_ASCII, 0, TIFF_SETGET_ASCII, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "OffsetTimeDigitized", NULL },
 	{ EXIFTAG_COMPONENTSCONFIGURATION, 4, 4, TIFF_UNDEFINED, 0, TIFF_SETGET_C0_UINT8, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "ComponentsConfiguration", NULL },
-	{ EXIFTAG_COMPRESSEDBITSPERPIXEL, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_DOUBLE, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "CompressedBitsPerPixel", NULL },
-	{ EXIFTAG_SHUTTERSPEEDVALUE, 1, 1, TIFF_SRATIONAL, 0, TIFF_SETGET_DOUBLE, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "ShutterSpeedValue", NULL },
-	{ EXIFTAG_APERTUREVALUE, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_DOUBLE, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "ApertureValue", NULL },
-	{ EXIFTAG_BRIGHTNESSVALUE, 1, 1, TIFF_SRATIONAL, 0, TIFF_SETGET_DOUBLE, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "BrightnessValue", NULL },
-	{ EXIFTAG_EXPOSUREBIASVALUE, 1, 1, TIFF_SRATIONAL, 0, TIFF_SETGET_DOUBLE, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "ExposureBiasValue", NULL },
-	{ EXIFTAG_MAXAPERTUREVALUE, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_DOUBLE, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "MaxApertureValue", NULL },
-	{ EXIFTAG_SUBJECTDISTANCE, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_DOUBLE, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "SubjectDistance", NULL },
+	{ EXIFTAG_COMPRESSEDBITSPERPIXEL, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "CompressedBitsPerPixel", NULL },
+	{ EXIFTAG_SHUTTERSPEEDVALUE, 1, 1, TIFF_SRATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "ShutterSpeedValue", NULL },
+	{ EXIFTAG_APERTUREVALUE, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "ApertureValue", NULL },
+	{ EXIFTAG_BRIGHTNESSVALUE, 1, 1, TIFF_SRATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "BrightnessValue", NULL },
+	{ EXIFTAG_EXPOSUREBIASVALUE, 1, 1, TIFF_SRATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "ExposureBiasValue", NULL },
+	{ EXIFTAG_MAXAPERTUREVALUE, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "MaxApertureValue", NULL },
+	/*--: EXIFTAG_SUBJECTDISTANCE: LibTiff returns value of "-1" if numerator equals 4294967295 (0xFFFFFFFF) to indicate infinite distance!
+	 * However, there are two other EXIF tags where numerator indicates a special value and six other cases where the denominator indicates special values,
+	 * which are not treated within LibTiff!! */
+	{ EXIFTAG_SUBJECTDISTANCE, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "SubjectDistance", NULL },
 	{ EXIFTAG_METERINGMODE, 1, 1, TIFF_SHORT, 0, TIFF_SETGET_UINT16, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "MeteringMode", NULL },
 	{ EXIFTAG_LIGHTSOURCE, 1, 1, TIFF_SHORT, 0, TIFF_SETGET_UINT16, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "LightSource", NULL },
 	{ EXIFTAG_FLASH, 1, 1, TIFF_SHORT, 0, TIFF_SETGET_UINT16, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "Flash", NULL },
-	{ EXIFTAG_FOCALLENGTH, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_DOUBLE, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "FocalLength", NULL },
+	{ EXIFTAG_FOCALLENGTH, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "FocalLength", NULL },
 	{ EXIFTAG_SUBJECTAREA, -1, -1, TIFF_SHORT, 0, TIFF_SETGET_C16_UINT16, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 1, "SubjectArea", NULL },
 	{ EXIFTAG_MAKERNOTE, -1, -1, TIFF_UNDEFINED, 0, TIFF_SETGET_C16_UINT8, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 1, "MakerNote", NULL },
 	{ EXIFTAG_USERCOMMENT, -1, -1, TIFF_UNDEFINED, 0, TIFF_SETGET_C16_UINT8, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 1, "UserComment", NULL },
 	{ EXIFTAG_SUBSECTIME, -1, -1, TIFF_ASCII, 0, TIFF_SETGET_ASCII, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "SubSecTime", NULL },
 	{ EXIFTAG_SUBSECTIMEORIGINAL, -1, -1, TIFF_ASCII, 0, TIFF_SETGET_ASCII, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "SubSecTimeOriginal", NULL },
 	{ EXIFTAG_SUBSECTIMEDIGITIZED, -1, -1, TIFF_ASCII, 0, TIFF_SETGET_ASCII, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "SubSecTimeDigitized", NULL },
+	{ EXIFTAG_TEMPERATURE, 1, 1, TIFF_SRATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "Temperature", NULL },
+	{ EXIFTAG_HUMIDITY, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "Humidity", NULL },
+	{ EXIFTAG_PRESSURE, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "Pressure", NULL },
+	{ EXIFTAG_WATERDEPTH, 1, 1, TIFF_SRATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "WaterDepth", NULL },
+	{ EXIFTAG_ACCELERATION, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "Acceleration", NULL },
+	{ EXIFTAG_CAMERAELEVATIONANGLE, 1, 1, TIFF_SRATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "CameraElevationAngle", NULL },
 	{ EXIFTAG_FLASHPIXVERSION, 4, 4, TIFF_UNDEFINED, 0, TIFF_SETGET_C0_UINT8, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "FlashpixVersion", NULL },
 	{ EXIFTAG_COLORSPACE, 1, 1, TIFF_SHORT, 0, TIFF_SETGET_UINT16, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "ColorSpace", NULL },
 	{ EXIFTAG_PIXELXDIMENSION, 1, 1, TIFF_LONG, 0, TIFF_SETGET_UINT32, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "PixelXDimension", NULL },
 	{ EXIFTAG_PIXELYDIMENSION, 1, 1, TIFF_LONG, 0, TIFF_SETGET_UINT32, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "PixelYDimension", NULL },
 	{ EXIFTAG_RELATEDSOUNDFILE, 13, 13, TIFF_ASCII, 0, TIFF_SETGET_ASCII, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "RelatedSoundFile", NULL },
-	{ EXIFTAG_FLASHENERGY, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_DOUBLE, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "FlashEnergy", NULL },
+	{ EXIFTAG_FLASHENERGY, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "FlashEnergy", NULL },
 	{ EXIFTAG_SPATIALFREQUENCYRESPONSE, -1, -1, TIFF_UNDEFINED, 0, TIFF_SETGET_C16_UINT8, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 1, "SpatialFrequencyResponse", NULL },
-	{ EXIFTAG_FOCALPLANEXRESOLUTION, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_DOUBLE, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "FocalPlaneXResolution", NULL },
-	{ EXIFTAG_FOCALPLANEYRESOLUTION, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_DOUBLE, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "FocalPlaneYResolution", NULL },
+	{ EXIFTAG_FOCALPLANEXRESOLUTION, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "FocalPlaneXResolution", NULL },
+	{ EXIFTAG_FOCALPLANEYRESOLUTION, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "FocalPlaneYResolution", NULL },
 	{ EXIFTAG_FOCALPLANERESOLUTIONUNIT, 1, 1, TIFF_SHORT, 0, TIFF_SETGET_UINT16, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "FocalPlaneResolutionUnit", NULL },
 	{ EXIFTAG_SUBJECTLOCATION, 2, 2, TIFF_SHORT, 0, TIFF_SETGET_C0_UINT16, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "SubjectLocation", NULL },
-	{ EXIFTAG_EXPOSUREINDEX, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_DOUBLE, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "ExposureIndex", NULL },
+	{ EXIFTAG_EXPOSUREINDEX, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "ExposureIndex", NULL },
 	{ EXIFTAG_SENSINGMETHOD, 1, 1, TIFF_SHORT, 0, TIFF_SETGET_UINT16, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "SensingMethod", NULL },
 	{ EXIFTAG_FILESOURCE, 1, 1, TIFF_UNDEFINED, 0, TIFF_SETGET_UINT8, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "FileSource", NULL },
 	{ EXIFTAG_SCENETYPE, 1, 1, TIFF_UNDEFINED, 0, TIFF_SETGET_UINT8, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "SceneType", NULL },
@@ -265,22 +300,79 @@ exifFields[] = {
 	{ EXIFTAG_CUSTOMRENDERED, 1, 1, TIFF_SHORT, 0, TIFF_SETGET_UINT16, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "CustomRendered", NULL },
 	{ EXIFTAG_EXPOSUREMODE, 1, 1, TIFF_SHORT, 0, TIFF_SETGET_UINT16, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "ExposureMode", NULL },
 	{ EXIFTAG_WHITEBALANCE, 1, 1, TIFF_SHORT, 0, TIFF_SETGET_UINT16, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "WhiteBalance", NULL },
-	{ EXIFTAG_DIGITALZOOMRATIO, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_DOUBLE, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "DigitalZoomRatio", NULL },
+	{ EXIFTAG_DIGITALZOOMRATIO, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "DigitalZoomRatio", NULL },
 	{ EXIFTAG_FOCALLENGTHIN35MMFILM, 1, 1, TIFF_SHORT, 0, TIFF_SETGET_UINT16, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "FocalLengthIn35mmFilm", NULL },
 	{ EXIFTAG_SCENECAPTURETYPE, 1, 1, TIFF_SHORT, 0, TIFF_SETGET_UINT16, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "SceneCaptureType", NULL },
-	{ EXIFTAG_GAINCONTROL, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_DOUBLE, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "GainControl", NULL },
+	{ EXIFTAG_GAINCONTROL, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "GainControl", NULL },
 	{ EXIFTAG_CONTRAST, 1, 1, TIFF_SHORT, 0, TIFF_SETGET_UINT16, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "Contrast", NULL },
 	{ EXIFTAG_SATURATION, 1, 1, TIFF_SHORT, 0, TIFF_SETGET_UINT16, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "Saturation", NULL },
 	{ EXIFTAG_SHARPNESS, 1, 1, TIFF_SHORT, 0, TIFF_SETGET_UINT16, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "Sharpness", NULL },
 	{ EXIFTAG_DEVICESETTINGDESCRIPTION, -1, -1, TIFF_UNDEFINED, 0, TIFF_SETGET_C16_UINT8, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 1, "DeviceSettingDescription", NULL },
 	{ EXIFTAG_SUBJECTDISTANCERANGE, 1, 1, TIFF_SHORT, 0, TIFF_SETGET_UINT16, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "SubjectDistanceRange", NULL },
-	{ EXIFTAG_IMAGEUNIQUEID, 33, 33, TIFF_ASCII, 0, TIFF_SETGET_ASCII, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "ImageUniqueID", NULL }
+	{ EXIFTAG_IMAGEUNIQUEID, 33, 33, TIFF_ASCII, 0, TIFF_SETGET_ASCII, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "ImageUniqueID", NULL },
+	{ EXIFTAG_CAMERAOWNERNAME, -1, -1, TIFF_ASCII, 0, TIFF_SETGET_ASCII, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "CameraOwnerName", NULL },
+	{ EXIFTAG_BODYSERIALNUMBER, -1, -1, TIFF_ASCII, 0, TIFF_SETGET_ASCII, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "BodySerialNumber", NULL },
+	{ EXIFTAG_LENSSPECIFICATION, 4, 4, TIFF_RATIONAL, 0, TIFF_SETGET_C0_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "LensSpecification", NULL },
+	{ EXIFTAG_LENSMAKE, -1, -1, TIFF_ASCII, 0, TIFF_SETGET_ASCII, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "LensMake", NULL },
+	{ EXIFTAG_LENSMODEL, -1, -1, TIFF_ASCII, 0, TIFF_SETGET_ASCII, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "LensModel", NULL },
+	{ EXIFTAG_LENSSERIALNUMBER, -1, -1, TIFF_ASCII, 0, TIFF_SETGET_ASCII, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "LensSerialNumber", NULL },
+	{ EXIFTAG_GAMMA, 1, 1, TIFF_RATIONAL, 0, TIFF_SETGET_FLOAT, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "Gamma", NULL },
+	{ EXIFTAG_COMPOSITEIMAGE, 1, 1, TIFF_SHORT, 0, TIFF_SETGET_UINT16, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "CompositeImage", NULL },
+	{ EXIFTAG_SOURCEIMAGENUMBEROFCOMPOSITEIMAGE, 2, 2, TIFF_SHORT, 0, TIFF_SETGET_C0_UINT16, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 0, "SourceImageNumberOfCompositeImage", NULL },
+	{ EXIFTAG_SOURCEEXPOSURETIMESOFCOMPOSITEIMAGE, -1, -1, TIFF_UNDEFINED, 0, TIFF_SETGET_C16_UINT8, TIFF_SETGET_UNDEFINED, FIELD_CUSTOM, 1, 1, "SourceExposureTimesOfCompositeImage", NULL }
+};
+/*
+ * EXIF-GPS tags  (Version 2.31, July 2016; nothing changed for version 2.32 May 2019)
+ */
+
+static TIFFField
+gpsFields[] = {
+	/*  For the GPS tag definitions in gpsFields[] the standard definition for Rationals is TIFF_SETGET_DOUBLE and TIFF_SETGET_C0_FLOAT.
+	 *-- ATTENTION: After the upgrade with Rational2Double, the GPSTAG values can now be written and also read in double precision!
+	 *              In order to achieve double precision for GPS tags:
+	 *              Standard definitions for GPSTAG is kept to TIFF_SETGET_DOUBLE 
+	 *              and TIFF_SETGET_C0_FLOAT is changed to TIFF_SETGET_C0_DOUBLE.
+	 */
+	{		GPSTAG_VERSIONID	, 4, 4, 	TIFF_BYTE	, 0, 	TIFF_SETGET_C0_UINT8	, TIFF_SETGET_UINT8	, FIELD_CUSTOM	, 1, 0, 	"VersionID", NULL },
+	{		GPSTAG_LATITUDEREF	, 2, 2,	TIFF_ASCII	, 0, 	TIFF_SETGET_ASCII	, TIFF_SETGET_UNDEFINED	, FIELD_CUSTOM	, 1, 0, 	"LatitudeRef", NULL },
+	{		GPSTAG_LATITUDE	, 3, 3,	TIFF_RATIONAL	, 0, 	TIFF_SETGET_C0_DOUBLE	, TIFF_SETGET_UNDEFINED	, FIELD_CUSTOM	, 1, 0, 	"Latitude", NULL },
+	{		GPSTAG_LONGITUDEREF	, 2, 2,	TIFF_ASCII	, 0, 	TIFF_SETGET_ASCII	, TIFF_SETGET_UNDEFINED	, FIELD_CUSTOM	, 1, 0, 	"LongitudeRef", NULL },
+	{		GPSTAG_LONGITUDE	, 3, 3,	TIFF_RATIONAL	, 0, 	TIFF_SETGET_C0_DOUBLE	, TIFF_SETGET_UNDEFINED	, FIELD_CUSTOM	, 1, 0, 	"Longitude", NULL },
+	{		GPSTAG_ALTITUDEREF	, 1, 1,	TIFF_BYTE	, 0, 	TIFF_SETGET_UINT8	, TIFF_SETGET_UNDEFINED	, FIELD_CUSTOM	, 1, 0, 	"AltitudeRef", NULL },
+	{		GPSTAG_ALTITUDE	, 1, 1,	TIFF_RATIONAL	, 0, 	TIFF_SETGET_DOUBLE	, TIFF_SETGET_UNDEFINED	, FIELD_CUSTOM	, 1, 0, 	"Altitude", NULL },
+	{		GPSTAG_TIMESTAMP	, 3, 3,	TIFF_RATIONAL	, 0, 	TIFF_SETGET_C0_DOUBLE	, TIFF_SETGET_UNDEFINED	, FIELD_CUSTOM	, 1, 0, 	"TimeStamp", NULL },
+	{		GPSTAG_SATELLITES	, -1, -1,	TIFF_ASCII	, 0, 	TIFF_SETGET_ASCII	, TIFF_SETGET_UNDEFINED	, FIELD_CUSTOM	, 1, 0, 	"Satellites", NULL },
+	{		GPSTAG_STATUS	, 2, 2,	TIFF_ASCII	, 0, 	TIFF_SETGET_ASCII	, TIFF_SETGET_UNDEFINED	, FIELD_CUSTOM	, 1, 0, 	"Status", NULL },
+	{		GPSTAG_MEASUREMODE	, 2, 2,	TIFF_ASCII	, 0, 	TIFF_SETGET_ASCII	, TIFF_SETGET_UNDEFINED	, FIELD_CUSTOM	, 1, 0, 	"MeasureMode", NULL },
+	{		GPSTAG_DOP	, 1, 1,	TIFF_RATIONAL	, 0, 	TIFF_SETGET_DOUBLE	, TIFF_SETGET_UNDEFINED	, FIELD_CUSTOM	, 1, 0, 	"DOP", NULL },
+	{		GPSTAG_SPEEDREF	, 2, 2,	TIFF_ASCII	, 0, 	TIFF_SETGET_ASCII	, TIFF_SETGET_UNDEFINED	, FIELD_CUSTOM	, 1, 0, 	"SpeedRef", NULL },
+	{		GPSTAG_SPEED	, 1, 1,	TIFF_RATIONAL	, 0, 	TIFF_SETGET_DOUBLE	, TIFF_SETGET_UNDEFINED	, FIELD_CUSTOM	, 1, 0, 	"Speed", NULL },
+	{		GPSTAG_TRACKREF	, 2, 2,	TIFF_ASCII	, 0, 	TIFF_SETGET_ASCII	, TIFF_SETGET_UNDEFINED	, FIELD_CUSTOM	, 1, 0, 	"TrackRef", NULL },
+	{		GPSTAG_TRACK	, 1, 1,	TIFF_RATIONAL	, 0, 	TIFF_SETGET_DOUBLE	, TIFF_SETGET_UNDEFINED	, FIELD_CUSTOM	, 1, 0, 	"Track", NULL },
+	{		GPSTAG_IMGDIRECTIONREF	, 2, 2,	TIFF_ASCII	, 0, 	TIFF_SETGET_ASCII	, TIFF_SETGET_UNDEFINED	, FIELD_CUSTOM	, 1, 0, 	"ImgDirectionRef", NULL },
+	{		GPSTAG_IMGDIRECTION	, 1, 1,	TIFF_RATIONAL	, 0, 	TIFF_SETGET_DOUBLE	, TIFF_SETGET_UNDEFINED	, FIELD_CUSTOM	, 1, 0, 	"ImgDirection", NULL },
+	{		GPSTAG_MAPDATUM	, -1, -1,	TIFF_ASCII	, 0, 	TIFF_SETGET_ASCII	, TIFF_SETGET_UNDEFINED	, FIELD_CUSTOM	, 1, 0, 	"MapDatum", NULL },
+	{		GPSTAG_DESTLATITUDEREF	, 2, 2,	TIFF_ASCII	, 0, 	TIFF_SETGET_ASCII	, TIFF_SETGET_UNDEFINED	, FIELD_CUSTOM	, 1, 0, 	"DestLatitudeRef", NULL },
+	{		GPSTAG_DESTLATITUDE	, 3, 3,	TIFF_RATIONAL	, 0, 	TIFF_SETGET_C0_DOUBLE	, TIFF_SETGET_UNDEFINED	, FIELD_CUSTOM	, 1, 0, 	"DestLatitude", NULL },
+	{		GPSTAG_DESTLONGITUDEREF	, 2, 2,	TIFF_ASCII	, 0, 	TIFF_SETGET_ASCII	, TIFF_SETGET_UNDEFINED	, FIELD_CUSTOM	, 1, 0, 	"DestLongitudeRef", NULL },
+	{		GPSTAG_DESTLONGITUDE	, 3, 3,	TIFF_RATIONAL	, 0, 	TIFF_SETGET_C0_DOUBLE	, TIFF_SETGET_UNDEFINED	, FIELD_CUSTOM	, 1, 0, 	"DestLongitude", NULL },
+	{		GPSTAG_DESTBEARINGREF	, 2, 2,	TIFF_ASCII	, 0, 	TIFF_SETGET_ASCII	, TIFF_SETGET_UNDEFINED	, FIELD_CUSTOM	, 1, 0, 	"DestBearingRef", NULL },
+	{		GPSTAG_DESTBEARING	, 1, 1,	TIFF_RATIONAL	, 0, 	TIFF_SETGET_DOUBLE	, TIFF_SETGET_UNDEFINED	, FIELD_CUSTOM	, 1, 0, 	"DestBearing", NULL },
+	{		GPSTAG_DESTDISTANCEREF	, 2, 2,	TIFF_ASCII	, 0, 	TIFF_SETGET_ASCII	, TIFF_SETGET_UNDEFINED	, FIELD_CUSTOM	, 1, 0, 	"DestDistanceRef", NULL },
+	{		GPSTAG_DESTDISTANCE	, 1, 1,	TIFF_RATIONAL	, 0, 	TIFF_SETGET_DOUBLE	, TIFF_SETGET_UNDEFINED	, FIELD_CUSTOM	, 1, 0, 	"DestDistance", NULL },
+	{		GPSTAG_PROCESSINGMETHOD	, -1, -1,	TIFF_UNDEFINED	, 0, 	TIFF_SETGET_C16_UINT8	, TIFF_SETGET_UNDEFINED	, FIELD_CUSTOM	, 1, 1, 	"ProcessingMethod", NULL },
+	{		GPSTAG_AREAINFORMATION	, -1, -1,	TIFF_UNDEFINED	, 0, 	TIFF_SETGET_C16_UINT8	, TIFF_SETGET_UNDEFINED	, FIELD_CUSTOM	, 1, 1, 	"AreaInformation", NULL },
+	{		GPSTAG_DATESTAMP	, 11, 11,	TIFF_ASCII	, 0, 	TIFF_SETGET_ASCII	, TIFF_SETGET_UNDEFINED	, FIELD_CUSTOM	, 1, 0, 	"DateStamp", NULL },
+	{		GPSTAG_DIFFERENTIAL	, 1, 1,	TIFF_SHORT	, 0, 	TIFF_SETGET_UINT16	, TIFF_SETGET_UNDEFINED	, FIELD_CUSTOM	, 1, 0, 	"Differential", NULL },
+	{		GPSTAG_GPSHPOSITIONINGERROR	, 1, 1,	TIFF_RATIONAL	, 0, 	TIFF_SETGET_DOUBLE	, TIFF_SETGET_UNDEFINED	, FIELD_CUSTOM	, 1, 0, 	"HorizontalPositioningError", NULL }
 };

 static const TIFFFieldArray
 tiffFieldArray = { tfiatImage, 0, TIFFArrayCount(tiffFields), (TIFFField*) tiffFields };
 static const TIFFFieldArray
 exifFieldArray = { tfiatExif, 0, TIFFArrayCount(exifFields), (TIFFField*) exifFields };
+static const TIFFFieldArray
+gpsFieldArray = { tfiatGps, 0, TIFFArrayCount(gpsFields), (TIFFField*) gpsFields };

 /*
 *  We have our own local lfind() equivalent to avoid subtle differences
@@ -313,6 +405,12 @@ _TIFFGetExifFields(void)
 	return(&exifFieldArray);
 }

+const TIFFFieldArray*
+_TIFFGetGpsFields(void)
+{
+	return(&gpsFieldArray);
+}
+
 void
 _TIFFSetupFields(TIFF* tif, const TIFFFieldArray* fieldarray)
 {
@@ -502,6 +600,82 @@ _TIFFDataSize(TIFFDataType type)
 	}
 }

+/*
+ * Rational2Double: 
+ * Return size of TIFFSetGetFieldType in bytes.
+ *
+ * XXX: TIFF_RATIONAL values for FIELD_CUSTOM are stored internally as 4-byte float.
+ * However, some of them should be stored internally as 8-byte double. 
+ * This is now managed by the SetGetField of the tag-definition!
+ */
+int
+_TIFFSetGetFieldSize(TIFFSetGetFieldType setgettype)
+{
+	switch (setgettype)
+	{
+		case TIFF_SETGET_UNDEFINED:
+		case TIFF_SETGET_ASCII:
+		case TIFF_SETGET_C0_ASCII:
+		case TIFF_SETGET_C16_ASCII:
+		case TIFF_SETGET_C32_ASCII:
+		case TIFF_SETGET_OTHER:
+		    return 0;
+		case TIFF_SETGET_UINT8:
+		case TIFF_SETGET_SINT8:
+		case TIFF_SETGET_C0_UINT8:
+		case TIFF_SETGET_C0_SINT8:
+		case TIFF_SETGET_C16_UINT8:
+		case TIFF_SETGET_C16_SINT8:
+		case TIFF_SETGET_C32_UINT8:
+		case TIFF_SETGET_C32_SINT8:
+		    return 1;
+		case TIFF_SETGET_UINT16:
+		case TIFF_SETGET_SINT16:
+		case TIFF_SETGET_C0_UINT16:
+		case TIFF_SETGET_C0_SINT16:
+		case TIFF_SETGET_C16_UINT16:
+		case TIFF_SETGET_C16_SINT16:
+		case TIFF_SETGET_C32_UINT16:
+		case TIFF_SETGET_C32_SINT16:
+		    return 2;
+		case TIFF_SETGET_INT:
+		case TIFF_SETGET_UINT32:
+		case TIFF_SETGET_SINT32:
+		case TIFF_SETGET_FLOAT:
+		case TIFF_SETGET_UINT16_PAIR:
+		case TIFF_SETGET_C0_UINT32:
+		case TIFF_SETGET_C0_SINT32:
+		case TIFF_SETGET_C0_FLOAT:
+		case TIFF_SETGET_C16_UINT32:
+		case TIFF_SETGET_C16_SINT32:
+		case TIFF_SETGET_C16_FLOAT:
+		case TIFF_SETGET_C32_UINT32:
+		case TIFF_SETGET_C32_SINT32:
+		case TIFF_SETGET_C32_FLOAT:
+		    return 4;
+		case TIFF_SETGET_UINT64:
+		case TIFF_SETGET_SINT64:
+		case TIFF_SETGET_DOUBLE:
+		case TIFF_SETGET_IFD8:
+		case TIFF_SETGET_C0_UINT64:
+		case TIFF_SETGET_C0_SINT64:
+		case TIFF_SETGET_C0_DOUBLE:
+		case TIFF_SETGET_C0_IFD8:
+		case TIFF_SETGET_C16_UINT64:
+		case TIFF_SETGET_C16_SINT64:
+		case TIFF_SETGET_C16_DOUBLE:
+		case TIFF_SETGET_C16_IFD8:
+		case TIFF_SETGET_C32_UINT64:
+		case TIFF_SETGET_C32_SINT64:
+		case TIFF_SETGET_C32_DOUBLE:
+		case TIFF_SETGET_C32_IFD8:
+		    return 8;
+		default:
+		    return 0;
+	}
+} /*-- _TIFFSetGetFieldSize --- */
+
+
 const TIFFField*
 TIFFFindField(TIFF* tif, uint32 tag, TIFFDataType dt)
 {
@@ -1062,10 +1236,6 @@ _TIFFCheckFieldIsValidForCodec(TIFF *tif, ttag_t tag)
 		if (tag == TIFFTAG_LERC_PARAMETERS)
 		    return 1;
 		break;
-		  case COMPRESSION_WEBP:
-		if (tag == TIFFTAG_PREDICTOR)
-				return 1;
-		break;
 	}
 	return 0;
 }
@@ -73,6 +73,7 @@ typedef struct {
 	int	EOLcnt;			/* count of EOL codes recognized */
 	TIFFFaxFillFunc fill;		/* fill routine */
 	uint32*	runs;			/* b&w runs for current/previous row */
+	uint32	nruns;			/* size of the refruns / curruns arrays */
 	uint32*	refruns;		/* runs for reference line */
 	uint32*	curruns;		/* runs for current line */

@@ -160,7 +161,9 @@ Fax3PreDecode(TIFF* tif, uint16 s)
 	 */
 	sp->bitmap =
 	    TIFFGetBitRevTable(tif->tif_dir.td_fillorder != FILLORDER_LSB2MSB);
+	sp->curruns = sp->runs;
 	if (sp->refruns) {		/* init reference line to white */
+		sp->refruns = sp->runs + sp->nruns;
 		sp->refruns[0] = (uint32) sp->b.rowpixels;
 		sp->refruns[1] = 0;
 	}
@@ -218,8 +221,12 @@ Fax3PrematureEOF(const char* module, TIFF* tif, uint32 line, uint32 a0)

 #define	Nop

-/*
+/**
 * Decode the requested amount of G3 1D-encoded data.
+ * @param buf destination buffer
+ * @param occ available bytes in destination buffer
+ * @param s number of planes (ignored)
+ * @returns 1 for success, -1 in case of error
 */
 static int
 Fax3Decode1D(TIFF* tif, uint8* buf, tmsize_t occ, uint16 s)
@@ -300,7 +307,9 @@ Fax3Decode2D(TIFF* tif, uint8* buf, tmsize_t occ, uint16 s)
 		else
 			EXPAND2D(EOF2Da);
 		(*sp->fill)(buf, thisrun, pa, lastx);
-		SETVALUE(0);		/* imaginary change for reference */
+		if (pa < thisrun + sp->nruns) {
+			SETVALUE(0);	/* imaginary change for reference */
+		}
 		SWAP(uint32*, sp->curruns, sp->refruns);
 		buf += sp->b.rowbytes;
 		occ -= sp->b.rowbytes;
@@ -506,7 +515,7 @@ Fax3SetupState(TIFF* tif)
 	int needsRefLine;
 	Fax3CodecState* dsp = (Fax3CodecState*) Fax3State(tif);
 	tmsize_t rowbytes;
-	uint32 rowpixels, nruns;
+	uint32 rowpixels;

 	if (td->td_bitspersample != 1) {
 		TIFFErrorExt(tif->tif_clientdata, module,
@@ -523,6 +532,13 @@ Fax3SetupState(TIFF* tif)
 		rowbytes = TIFFScanlineSize(tif);
 		rowpixels = td->td_imagewidth;
 	}
+	if ((uint64)rowbytes < ((uint64)rowpixels + 7) / 8)
+	{
+		TIFFErrorExt(tif->tif_clientdata, module,
+			"Inconsistent number of bytes per row : rowbytes=%lu rowpixels=%lu",
+			(unsigned long)(rowbytes), (unsigned long)(rowpixels));
+		return (0);
+	}
 	sp->rowbytes = rowbytes;
 	sp->rowpixels = rowpixels;
 	/*
@@ -539,26 +555,26 @@ Fax3SetupState(TIFF* tif)
 	  TIFFroundup and TIFFSafeMultiply return zero on integer overflow
 	*/
 	dsp->runs=(uint32*) NULL;
-	nruns = TIFFroundup_32(rowpixels,32);
+	dsp->nruns = TIFFroundup_32(rowpixels,32);
 	if (needsRefLine) {
-		nruns = TIFFSafeMultiply(uint32,nruns,2);
+		dsp->nruns = TIFFSafeMultiply(uint32,dsp->nruns,2);
 	}
-	if ((nruns == 0) || (TIFFSafeMultiply(uint32,nruns,2) == 0)) {
+	if ((dsp->nruns == 0) || (TIFFSafeMultiply(uint32,dsp->nruns,2) == 0)) {
 		TIFFErrorExt(tif->tif_clientdata, tif->tif_name,
 			     "Row pixels integer overflow (rowpixels %u)",
 			     rowpixels);
 		return (0);
 	}
 	dsp->runs = (uint32*) _TIFFCheckMalloc(tif,
-					       TIFFSafeMultiply(uint32,nruns,2),
+					       TIFFSafeMultiply(uint32,dsp->nruns,2),
 					       sizeof (uint32),
 					       "for Group 3/4 run arrays");
 	if (dsp->runs == NULL)
 		return (0);
-	memset( dsp->runs, 0, TIFFSafeMultiply(uint32,nruns,2)*sizeof(uint32));
+	memset( dsp->runs, 0, TIFFSafeMultiply(uint32,dsp->nruns,2)*sizeof(uint32));
 	dsp->curruns = dsp->runs;
 	if (needsRefLine)
-		dsp->refruns = dsp->runs + nruns;
+		dsp->refruns = dsp->runs + dsp->nruns;
 	else
 		dsp->refruns = NULL;
 	if (td->td_compression == COMPRESSION_CCITTFAX3
@@ -594,15 +610,19 @@ Fax3SetupState(TIFF* tif)
 */

 #define	Fax3FlushBits(tif, sp) {				\
-	if ((tif)->tif_rawcc >= (tif)->tif_rawdatasize)		\
-		(void) TIFFFlushData1(tif);			\
+	if ((tif)->tif_rawcc >= (tif)->tif_rawdatasize) {	\
+		if( !TIFFFlushData1(tif) )			\
+			return 0;				\
+        }							\
 	*(tif)->tif_rawcp++ = (uint8) (sp)->data;		\
 	(tif)->tif_rawcc++;					\
 	(sp)->data = 0, (sp)->bit = 8;				\
 }
 #define	_FlushBits(tif) {					\
-	if ((tif)->tif_rawcc >= (tif)->tif_rawdatasize)		\
-		(void) TIFFFlushData1(tif);			\
+	if ((tif)->tif_rawcc >= (tif)->tif_rawdatasize) {	\
+		if( !TIFFFlushData1(tif) )			\
+			return 0;				\
+        }							\
 	*(tif)->tif_rawcp++ = (uint8) data;		\
 	(tif)->tif_rawcc++;					\
 	data = 0, bit = 8;					\
@@ -627,7 +647,7 @@ static const int _msbmask[9] =
 * the output stream.  Values are
 * assumed to be at most 16 bits.
 */
-static void
+static int
 Fax3PutBits(TIFF* tif, unsigned int bits, unsigned int length)
 {
 	Fax3CodecState* sp = EncoderState(tif);
@@ -638,6 +658,7 @@ Fax3PutBits(TIFF* tif, unsigned int bits, unsigned int length)

 	sp->data = data;
 	sp->bit = bit;
+        return 1;
 }

 /*
@@ -662,7 +683,7 @@ Fax3PutBits(TIFF* tif, unsigned int bits, unsigned int length)
 * appropriate table that holds the make-up and
 * terminating codes is supplied.
 */
-static void
+static int
 putspan(TIFF* tif, int32 span, const tableentry* tab)
 {
 	Fax3CodecState* sp = EncoderState(tif);
@@ -700,6 +721,8 @@ putspan(TIFF* tif, int32 span, const tableentry* tab)

 	sp->data = data;
 	sp->bit = bit;
+
+        return 1;
 }

 /*
@@ -708,7 +731,7 @@ putspan(TIFF* tif, int32 span, const tableentry* tab)
 * here.  We also handle writing the tag bit for the next
 * scanline when doing 2d encoding.
 */
-static void
+static int
 Fax3PutEOL(TIFF* tif)
 {
 	Fax3CodecState* sp = EncoderState(tif);
@@ -742,6 +765,8 @@ Fax3PutEOL(TIFF* tif)

 	sp->data = data;
 	sp->bit = bit;
+
+        return 1;
 }

 /*
@@ -991,12 +1016,14 @@ Fax3Encode1DRow(TIFF* tif, unsigned char* bp, uint32 bits)

 	for (;;) {
 		span = find0span(bp, bs, bits);		/* white span */
-		putspan(tif, span, TIFFFaxWhiteCodes);
+		if( !putspan(tif, span, TIFFFaxWhiteCodes) )
+                    return 0;
 		bs += span;
 		if (bs >= bits)
 			break;
 		span = find1span(bp, bs, bits);		/* black span */
-		putspan(tif, span, TIFFFaxBlackCodes);
+		if( !putspan(tif, span, TIFFFaxBlackCodes) )
+                    return 0;
 		bs += span;
 		if (bs >= bits)
 			break;
@@ -1048,21 +1075,28 @@ Fax3Encode2DRow(TIFF* tif, unsigned char* bp, unsigned char* rp, uint32 bits)
 			          (b1 < a1 && a1 - b1 <= 3U) ? -(int32)(a1 - b1) : 0x7FFFFFFF;
 			if (!(-3 <= d && d <= 3)) {	/* horizontal mode */
 				a2 = finddiff2(bp, a1, bits, PIXEL(bp,a1));
-				putcode(tif, &horizcode);
+				if( !putcode(tif, &horizcode) )
+                                    return 0;
 				if (a0+a1 == 0 || PIXEL(bp, a0) == 0) {
-					putspan(tif, a1-a0, TIFFFaxWhiteCodes);
-					putspan(tif, a2-a1, TIFFFaxBlackCodes);
+					if( !putspan(tif, a1-a0, TIFFFaxWhiteCodes) )
+                                            return 0;
+					if( !putspan(tif, a2-a1, TIFFFaxBlackCodes) )
+                                            return 0;
 				} else {
-					putspan(tif, a1-a0, TIFFFaxBlackCodes);
-					putspan(tif, a2-a1, TIFFFaxWhiteCodes);
+					if( !putspan(tif, a1-a0, TIFFFaxBlackCodes) )
+                                            return 0;
+					if( !putspan(tif, a2-a1, TIFFFaxWhiteCodes) )
+                                            return 0;
 				}
 				a0 = a2;
 			} else {			/* vertical mode */
-				putcode(tif, &vcodes[d+3]);
+				if( !putcode(tif, &vcodes[d+3]) )
+                                    return 0;
 				a0 = a1;
 			}
 		} else {				/* pass mode */
-			putcode(tif, &passcode);
+			if( !putcode(tif, &passcode) )
+                            return 0;
 			a0 = b2;
 		}
 		if (a0 >= bits)
@@ -1091,7 +1125,10 @@ Fax3Encode(TIFF* tif, uint8* bp, tmsize_t cc, uint16 s)
 	}
 	while (cc > 0) {
 		if ((sp->b.mode & FAXMODE_NOEOL) == 0)
-			Fax3PutEOL(tif);
+                {
+			if( !Fax3PutEOL(tif) )
+                            return 0;
+                }
 		if (is2DEncoding(sp)) {
 			if (sp->tag == G3_1D) {
 				if (!Fax3Encode1DRow(tif, bp, sp->b.rowpixels))
@@ -1128,8 +1165,8 @@ Fax3PostEncode(TIFF* tif)
 	return (1);
 }

-static void
-Fax3Close(TIFF* tif)
+static int
+_Fax3Close(TIFF* tif)
 {
 	if ((Fax3State(tif)->mode & FAXMODE_NORTC) == 0 && tif->tif_rawcp) {
 		Fax3CodecState* sp = EncoderState(tif);
@@ -1145,6 +1182,13 @@ Fax3Close(TIFF* tif)
 			Fax3PutBits(tif, code, length);
 		Fax3FlushBits(tif, sp);
 	}
+	return 1;
+}
+
+static void
+Fax3Close(TIFF* tif)
+{
+    _Fax3Close(tif);
 }

 static void
@@ -1453,6 +1497,13 @@ Fax4Decode(TIFF* tif, uint8* buf, tmsize_t occ, uint16 s)
 		EXPAND2D(EOFG4);
                if (EOLcnt)
                    goto EOFG4;
+		if (((lastx + 7) >> 3) > (int)occ)	/* check for buffer overrun */
+		{
+			TIFFErrorExt(tif->tif_clientdata, module,
+			             "Buffer overrun detected : %d bytes available, %d bits needed",
+			             (int)occ, lastx);
+			return -1;
+		}
 		(*sp->fill)(buf, thisrun, pa, lastx);
 		SETVALUE(0);		/* imaginary change for reference */
 		SWAP(uint32*, sp->curruns, sp->refruns);
@@ -1468,6 +1519,13 @@ Fax4Decode(TIFF* tif, uint8* buf, tmsize_t occ, uint16 s)
                    fputs( "Bad EOFB\n", stderr );
 #endif                
                ClrBits( 13 );
+		if (((lastx + 7) >> 3) > (int)occ)	/* check for buffer overrun */
+		{
+			TIFFErrorExt(tif->tif_clientdata, module,
+			             "Buffer overrun detected : %d bytes available, %d bits needed",
+			             (int)occ, lastx);
+			return -1;
+		}
 		(*sp->fill)(buf, thisrun, pa, lastx);
 		UNCACHE_STATE(tif, sp);
 		return ( sp->line ? 1 : -1);	/* don't error on badly-terminated strips */
@@ -240,6 +240,11 @@ static const char* StateNames[] = {
 * current row and reset decoding state.
 */
 #define SETVALUE(x) do {							\
+    if (pa >= thisrun + sp->nruns) {					\
+        TIFFErrorExt(tif->tif_clientdata, module, "Buffer overflow at line %u of %s %u",	\
+                    sp->line, isTiled(tif) ? "tile" : "strip", isTiled(tif) ? tif->tif_curtile : tif->tif_curstrip);	\
+        return (-1);							\
+    }									\
    *pa++ = RunLength + (x);						\
    a0 += (x);								\
    RunLength = 0;							\
@@ -377,6 +382,11 @@ done1d:									\
 */
 #define CHECK_b1 do {							\
    if (pa != thisrun) while (b1 <= a0 && b1 < lastx) {			\
+	if( pb + 1 >= sp->refruns + sp->nruns) { 			\
+	    TIFFErrorExt(tif->tif_clientdata, module, "Buffer overflow at line %u of %s %u",	\
+	                sp->line, isTiled(tif) ? "tile" : "strip", isTiled(tif) ? tif->tif_curtile : tif->tif_curstrip);	\
+	    return (-1);						\
+	}								\
 	b1 += pb[0] + pb[1];						\
 	pb += 2;							\
    }									\
@@ -387,10 +397,20 @@ done1d:									\
 */
 #define EXPAND2D(eoflab) do {						\
    while (a0 < lastx) {						\
+	if (pa >= thisrun + sp->nruns) {				\
+		TIFFErrorExt(tif->tif_clientdata, module, "Buffer overflow at line %u of %s %u",	\
+		             sp->line, isTiled(tif) ? "tile" : "strip", isTiled(tif) ? tif->tif_curtile : tif->tif_curstrip);	\
+		return (-1);						\
+	}								\
 	LOOKUP8(7, TIFFFaxMainTable, eof2d);				\
 	switch (TabEnt->State) {					\
 	case S_Pass:							\
 	    CHECK_b1;							\
+	    if( pb + 1 >= sp->refruns + sp->nruns) { 			\
+	        TIFFErrorExt(tif->tif_clientdata, module, "Buffer overflow at line %u of %s %u",	\
+	                sp->line, isTiled(tif) ? "tile" : "strip", isTiled(tif) ? tif->tif_curtile : tif->tif_curstrip);	\
+	        return (-1);						\
+	    }								\
 	    b1 += *pb++;						\
 	    RunLength += b1 - a0;					\
 	    a0 = b1;							\
@@ -469,20 +489,28 @@ done1d:									\
 	case S_V0:							\
 	    CHECK_b1;							\
 	    SETVALUE(b1 - a0);						\
+	    if( pb >= sp->refruns + sp->nruns) { 			\
+	        TIFFErrorExt(tif->tif_clientdata, module, "Buffer overflow at line %u of %s %u",	\
+	                sp->line, isTiled(tif) ? "tile" : "strip", isTiled(tif) ? tif->tif_curtile : tif->tif_curstrip);	\
+	        return (-1);						\
+	    }								\
 	    b1 += *pb++;						\
 	    break;							\
 	case S_VR:							\
 	    CHECK_b1;							\
 	    SETVALUE(b1 - a0 + TabEnt->Param);				\
+	    if( pb >= sp->refruns + sp->nruns) { 			\
+	        TIFFErrorExt(tif->tif_clientdata, module, "Buffer overflow at line %u of %s %u",	\
+	                sp->line, isTiled(tif) ? "tile" : "strip", isTiled(tif) ? tif->tif_curtile : tif->tif_curstrip);	\
+	        return (-1);						\
+	    }								\
 	    b1 += *pb++;						\
 	    break;							\
 	case S_VL:							\
 	    CHECK_b1;							\
-	    if (b1 <= (int) (a0 + TabEnt->Param)) {			\
-		if (b1 < (int) (a0 + TabEnt->Param) || pa != thisrun) {	\
-		    unexpected("VL", a0);				\
-		    goto eol2d;						\
-		}							\
+	    if (b1 < (int) (a0 + TabEnt->Param)) {			\
+		unexpected("VL", a0);				\
+		goto eol2d;						\
 	    }								\
 	    SETVALUE(b1 - a0 - TabEnt->Param);				\
 	    b1 -= *--pb;						\
@@ -529,6 +557,7 @@ eol2d:									\
    CLEANUP_RUNS();							\
 } while (0)
 #endif /* _FAX3_ */
+/* vim: set ts=8 sts=4 sw=4 noet: */
 /*
 * Local Variables:
 * mode: c
@@ -1,5 +1,6 @@
 /* WARNING, this file was automatically generated by the
    mkg3states program */
+#include <stdint.h>
 #include "tiff.h"
 #include "tif_fax3.h"
 const TIFFFaxTabEnt TIFFFaxMainTable[128] = {
@@ -45,36 +45,8 @@ TIFFFlush(TIFF* tif)
        && !(tif->tif_flags & TIFF_DIRTYDIRECT) 
        && tif->tif_mode == O_RDWR )
    {
-        uint64  *offsets=NULL, *sizes=NULL;
-
-        if( TIFFIsTiled(tif) )
-        {
-            if( TIFFGetField( tif, TIFFTAG_TILEOFFSETS, &offsets ) 
-                && TIFFGetField( tif, TIFFTAG_TILEBYTECOUNTS, &sizes ) 
-                && _TIFFRewriteField( tif, TIFFTAG_TILEOFFSETS, TIFF_LONG8, 
-                                      tif->tif_dir.td_nstrips, offsets )
-                && _TIFFRewriteField( tif, TIFFTAG_TILEBYTECOUNTS, TIFF_LONG8, 
-                                      tif->tif_dir.td_nstrips, sizes ) )
-            {
-                tif->tif_flags &= ~TIFF_DIRTYSTRIP;
-                tif->tif_flags &= ~TIFF_BEENWRITING;
-                return 1;
-            }
-        }
-        else
-        {
-            if( TIFFGetField( tif, TIFFTAG_STRIPOFFSETS, &offsets ) 
-                && TIFFGetField( tif, TIFFTAG_STRIPBYTECOUNTS, &sizes ) 
-                && _TIFFRewriteField( tif, TIFFTAG_STRIPOFFSETS, TIFF_LONG8, 
-                                      tif->tif_dir.td_nstrips, offsets )
-                && _TIFFRewriteField( tif, TIFFTAG_STRIPBYTECOUNTS, TIFF_LONG8, 
-                                      tif->tif_dir.td_nstrips, sizes ) )
-            {
-                tif->tif_flags &= ~TIFF_DIRTYSTRIP;
-                tif->tif_flags &= ~TIFF_BEENWRITING;
-                return 1;
-            }
-        }
+        if( TIFFForceStrileArrayWriting(tif) )
+            return 1;
    }

    if ((tif->tif_flags & (TIFF_DIRTYDIRECT|TIFF_DIRTYSTRIP)) 
@@ -84,6 +56,92 @@ TIFFFlush(TIFF* tif)
    return (1);
 }

+/*
+ * This is an advanced writing function that must be used in a particular
+ * sequence, and together with TIFFDeferStrileArrayWriting(),
+ * to make its intended effect. Its aim is to force the writing of
+ * the [Strip/Tile][Offsets/ByteCounts] arrays at the end of the file, when
+ * they have not yet been rewritten.
+ *
+ * The typical sequence of calls is:
+ * TIFFOpen()
+ * [ TIFFCreateDirectory(tif) ]
+ * Set fields with calls to TIFFSetField(tif, ...)
+ * TIFFDeferStrileArrayWriting(tif)
+ * TIFFWriteCheck(tif, ...)
+ * TIFFWriteDirectory(tif)
+ * ... potentially create other directories and come back to the above directory
+ * TIFFForceStrileArrayWriting(tif)
+ *
+ * Returns 1 in case of success, 0 otherwise.
+ */
+int TIFFForceStrileArrayWriting(TIFF* tif)
+{
+    static const char module[] = "TIFFForceStrileArrayWriting";
+    const int isTiled = TIFFIsTiled(tif);
+
+    if (tif->tif_mode == O_RDONLY)
+    {
+        TIFFErrorExt(tif->tif_clientdata, tif->tif_name,
+                     "File opened in read-only mode");
+        return 0;
+    }
+    if( tif->tif_diroff == 0 )
+    {
+        TIFFErrorExt(tif->tif_clientdata, module,
+                     "Directory has not yet been written");
+        return 0;
+    }
+    if( (tif->tif_flags & TIFF_DIRTYDIRECT) != 0 )
+    {
+        TIFFErrorExt(tif->tif_clientdata, module,
+                     "Directory has changes other than the strile arrays. "
+                     "TIFFRewriteDirectory() should be called instead");
+        return 0;
+    }
+
+    if( !(tif->tif_flags & TIFF_DIRTYSTRIP) )
+    {
+        if( !(tif->tif_dir.td_stripoffset_entry.tdir_tag != 0 &&
+             tif->tif_dir.td_stripoffset_entry.tdir_count == 0 &&
+             tif->tif_dir.td_stripoffset_entry.tdir_type == 0 &&
+             tif->tif_dir.td_stripoffset_entry.tdir_offset.toff_long8 == 0 &&
+             tif->tif_dir.td_stripbytecount_entry.tdir_tag != 0 &&
+             tif->tif_dir.td_stripbytecount_entry.tdir_count == 0 &&
+             tif->tif_dir.td_stripbytecount_entry.tdir_type == 0 &&
+             tif->tif_dir.td_stripbytecount_entry.tdir_offset.toff_long8 == 0) )
+        {
+            TIFFErrorExt(tif->tif_clientdata, module,
+                        "Function not called together with "
+                        "TIFFDeferStrileArrayWriting()");
+            return 0;
+        }
+
+        if (tif->tif_dir.td_stripoffset_p == NULL && !TIFFSetupStrips(tif))
+            return 0;
+    }
+
+    if( _TIFFRewriteField( tif,
+                           isTiled ? TIFFTAG_TILEOFFSETS :
+                                     TIFFTAG_STRIPOFFSETS,
+                           TIFF_LONG8,
+                           tif->tif_dir.td_nstrips,
+                           tif->tif_dir.td_stripoffset_p )
+        && _TIFFRewriteField( tif,
+                              isTiled ? TIFFTAG_TILEBYTECOUNTS :
+                                        TIFFTAG_STRIPBYTECOUNTS,
+                              TIFF_LONG8,
+                              tif->tif_dir.td_nstrips,
+                              tif->tif_dir.td_stripbytecount_p ) )
+    {
+        tif->tif_flags &= ~TIFF_DIRTYSTRIP;
+        tif->tif_flags &= ~TIFF_BEENWRITING;
+        return 1;
+    }
+
+    return 0;
+}
+
 /*
 * Flush buffered data to the file.
 *
@@ -29,6 +29,7 @@
 */
 #include "tiffiop.h"
 #include <stdio.h>
+#include <limits.h>

 static int gtTileContig(TIFFRGBAImage*, uint32*, uint32, uint32);
 static int gtTileSeparate(TIFFRGBAImage*, uint32*, uint32, uint32);
@@ -645,12 +646,20 @@ gtTileContig(TIFFRGBAImage* img, uint32* raster, uint32 w, uint32 h)

    flip = setorientation(img);
    if (flip & FLIP_VERTICALLY) {
-	    y = h - 1;
-	    toskew = -(int32)(tw + w);
+        if ((tw + w) > INT_MAX) {
+            TIFFErrorExt(tif->tif_clientdata, TIFFFileName(tif), "%s", "unsupported tile size (too wide)");
+            return (0);
+        }
+        y = h - 1;
+        toskew = -(int32)(tw + w);
    }
    else {
-	    y = 0;
-	    toskew = -(int32)(tw - w);
+        if (tw > (INT_MAX + w)) {
+            TIFFErrorExt(tif->tif_clientdata, TIFFFileName(tif), "%s", "unsupported tile size (too wide)");
+            return (0);
+        }
+        y = 0;
+        toskew = -(int32)(tw - w);
    }
     
    /*
@@ -755,9 +764,8 @@ gtTileSeparate(TIFFRGBAImage* img, uint32* raster, uint32 w, uint32 h)
 	uint32 leftmost_tw;

 	tilesize = TIFFTileSize(tif);  
-	bufsize = TIFFSafeMultiply(tmsize_t,alpha?4:3,tilesize);
+	bufsize = _TIFFMultiplySSize(tif, alpha?4:3,tilesize, "gtTileSeparate");
 	if (bufsize == 0) {
-		TIFFErrorExt(tif->tif_clientdata, TIFFFileName(tif), "Integer overflow in %s", "gtTileSeparate");
 		return (0);
 	}

@@ -766,10 +774,18 @@ gtTileSeparate(TIFFRGBAImage* img, uint32* raster, uint32 w, uint32 h)

 	flip = setorientation(img);
 	if (flip & FLIP_VERTICALLY) {
+		if ((tw + w) > INT_MAX) {
+            TIFFErrorExt(tif->tif_clientdata, TIFFFileName(tif), "%s", "unsupported tile size (too wide)");
+            return (0);
+        }
 		y = h - 1;
 		toskew = -(int32)(tw + w);
 	}
 	else {
+		if (tw > (INT_MAX + w)) {
+            TIFFErrorExt(tif->tif_clientdata, TIFFFileName(tif), "%s", "unsupported tile size (too wide)");
+            return (0);
+        }
 		y = 0;
 		toskew = -(int32)(tw - w);
 	}
@@ -937,6 +953,10 @@ gtStripContig(TIFFRGBAImage* img, uint32* raster, uint32 w, uint32 h)

 	flip = setorientation(img);
 	if (flip & FLIP_VERTICALLY) {
+		if ( w > INT_MAX ) {
+        	TIFFErrorExt(tif->tif_clientdata, TIFFFileName(tif), "Width overflow");
+			return (0);
+		}
 		y = h - 1;
 		toskew = -(int32)(w + w);
 	} else {
@@ -950,16 +970,23 @@ gtStripContig(TIFFRGBAImage* img, uint32* raster, uint32 w, uint32 h)
 	fromskew = (w < imagewidth ? imagewidth - w : 0);
 	for (row = 0; row < h; row += nrow)
 	{
+		uint32 temp;
 		rowstoread = rowsperstrip - (row + img->row_offset) % rowsperstrip;
 		nrow = (row + rowstoread > h ? h - row : rowstoread);
 		nrowsub = nrow;
 		if ((nrowsub%subsamplingver)!=0)
 			nrowsub+=subsamplingver-nrowsub%subsamplingver;
+		temp = (row + img->row_offset)%rowsperstrip + nrowsub;
+		if( scanline > 0 && temp > (size_t)(TIFF_TMSIZE_T_MAX / scanline) )
+		{
+			TIFFErrorExt(tif->tif_clientdata, TIFFFileName(tif), "Integer overflow in gtStripContig");
+			return 0;
+		}
 		if (_TIFFReadEncodedStripAndAllocBuffer(tif,
 		    TIFFComputeStrip(tif,row+img->row_offset, 0),
 		    (void**)(&buf),
                    maxstripsize,
-		    ((row + img->row_offset)%rowsperstrip + nrowsub) * scanline)==(tmsize_t)(-1)
+		    temp * scanline)==(tmsize_t)(-1)
 		    && (buf == NULL || img->stoponerr))
 		{
 			ret = 0;
@@ -1019,14 +1046,17 @@ gtStripSeparate(TIFFRGBAImage* img, uint32* raster, uint32 w, uint32 h)
        uint16 colorchannels;

 	stripsize = TIFFStripSize(tif);  
-	bufsize = TIFFSafeMultiply(tmsize_t,alpha?4:3,stripsize);
+	bufsize = _TIFFMultiplySSize(tif,alpha?4:3,stripsize, "gtStripSeparate");
 	if (bufsize == 0) {
-		TIFFErrorExt(tif->tif_clientdata, TIFFFileName(tif), "Integer overflow in %s", "gtStripSeparate");
 		return (0);
 	}

 	flip = setorientation(img);
 	if (flip & FLIP_VERTICALLY) {
+		if ( w > INT_MAX ) {
+        	TIFFErrorExt(tif->tif_clientdata, TIFFFileName(tif), "Width overflow");
+			return (0);
+		}
 		y = h - 1;
 		toskew = -(int32)(w + w);
 	}
@@ -1053,15 +1083,22 @@ gtStripSeparate(TIFFRGBAImage* img, uint32* raster, uint32 w, uint32 h)
 	fromskew = (w < imagewidth ? imagewidth - w : 0);
 	for (row = 0; row < h; row += nrow)
 	{
+                uint32 temp;
 		rowstoread = rowsperstrip - (row + img->row_offset) % rowsperstrip;
 		nrow = (row + rowstoread > h ? h - row : rowstoread);
 		offset_row = row + img->row_offset;
+                temp = (row + img->row_offset)%rowsperstrip + nrow;
+                if( scanline > 0 && temp > (size_t)(TIFF_TMSIZE_T_MAX / scanline) )
+                {
+                        TIFFErrorExt(tif->tif_clientdata, TIFFFileName(tif), "Integer overflow in gtStripSeparate");
+                        return 0;
+                }
                if( buf == NULL )
                {
                    if (_TIFFReadEncodedStripAndAllocBuffer(
                            tif, TIFFComputeStrip(tif, offset_row, 0),
                            (void**) &buf, bufsize,
-                            ((row + img->row_offset)%rowsperstrip + nrow) * scanline)==(tmsize_t)(-1)
+                            temp * scanline)==(tmsize_t)(-1)
                        && (buf == NULL || img->stoponerr))
                    {
                            ret = 0;
@@ -1081,7 +1118,7 @@ gtStripSeparate(TIFFRGBAImage* img, uint32* raster, uint32 w, uint32 h)
                    }
                }
 		else if (TIFFReadEncodedStrip(tif, TIFFComputeStrip(tif, offset_row, 0),
-		    p0, ((row + img->row_offset)%rowsperstrip + nrow) * scanline)==(tmsize_t)(-1)
+		    p0, temp * scanline)==(tmsize_t)(-1)
 		    && img->stoponerr)
 		{
 			ret = 0;
@@ -1089,7 +1126,7 @@ gtStripSeparate(TIFFRGBAImage* img, uint32* raster, uint32 w, uint32 h)
 		}
 		if (colorchannels > 1 
                    && TIFFReadEncodedStrip(tif, TIFFComputeStrip(tif, offset_row, 1),
-                                            p1, ((row + img->row_offset)%rowsperstrip + nrow) * scanline) == (tmsize_t)(-1)
+                                            p1, temp * scanline) == (tmsize_t)(-1)
 		    && img->stoponerr)
 		{
 			ret = 0;
@@ -1097,7 +1134,7 @@ gtStripSeparate(TIFFRGBAImage* img, uint32* raster, uint32 w, uint32 h)
 		}
 		if (colorchannels > 1 
                    && TIFFReadEncodedStrip(tif, TIFFComputeStrip(tif, offset_row, 2),
-                                            p2, ((row + img->row_offset)%rowsperstrip + nrow) * scanline) == (tmsize_t)(-1)
+                                            p2, temp * scanline) == (tmsize_t)(-1)
 		    && img->stoponerr)
 		{
 			ret = 0;
@@ -1106,7 +1143,7 @@ gtStripSeparate(TIFFRGBAImage* img, uint32* raster, uint32 w, uint32 h)
 		if (alpha)
 		{
 			if (TIFFReadEncodedStrip(tif, TIFFComputeStrip(tif, offset_row, colorchannels),
-			    pa, ((row + img->row_offset)%rowsperstrip + nrow) * scanline)==(tmsize_t)(-1)
+			    pa, temp * scanline)==(tmsize_t)(-1)
 			    && img->stoponerr)
 			{
 				ret = 0;
@@ -2957,7 +2994,7 @@ TIFFReadRGBATileExt(TIFF* tif, uint32 col, uint32 row, uint32 * raster, int stop
    if( !TIFFIsTiled( tif ) )
    {
 		TIFFErrorExt(tif->tif_clientdata, TIFFFileName(tif),
-				  "Can't use TIFFReadRGBATile() with stripped file.");
+				  "Can't use TIFFReadRGBATile() with striped file.");
 		return (0);
    }
    
@@ -199,6 +199,7 @@ static int JBIGEncode(TIFF* tif, uint8* buffer, tmsize_t size, uint16 s)

 int TIFFInitJBIG(TIFF* tif, int scheme)
 {
+        (void)scheme;
 	assert(scheme == COMPRESSION_JBIG);

 	/*
@@ -466,7 +466,8 @@ std_empty_output_buffer(j_compress_ptr cinfo)
       }
 #endif

-	TIFFFlushData1(tif);
+	if( !TIFFFlushData1(tif) )
+            return FALSE;
 	sp->dest.next_output_byte = (JOCTET*) tif->tif_rawdata;
 	sp->dest.free_in_buffer = (size_t) tif->tif_rawdatasize;

@@ -780,12 +781,9 @@ JPEGFixupTagsSubsampling(TIFF* tif)
 	 */
 	static const char module[] = "JPEGFixupTagsSubsampling";
 	struct JPEGFixupTagsSubsamplingData m;
+        uint64 fileoffset = TIFFGetStrileOffset(tif, 0);

-        _TIFFFillStriles( tif );
-        
-        if( tif->tif_dir.td_stripbytecount == NULL
-            || tif->tif_dir.td_stripoffset == NULL
-            || tif->tif_dir.td_stripbytecount[0] == 0 )
+        if( fileoffset == 0 )
        {
            /* Do not even try to check if the first strip/tile does not
               yet exist, as occurs when GDAL has created a new NULL file
@@ -804,9 +802,9 @@ JPEGFixupTagsSubsampling(TIFF* tif)
 	}
 	m.buffercurrentbyte=NULL;
 	m.bufferbytesleft=0;
-	m.fileoffset=tif->tif_dir.td_stripoffset[0];
+	m.fileoffset=fileoffset;
 	m.filepositioned=0;
-	m.filebytesleft=tif->tif_dir.td_stripbytecount[0];
+	m.filebytesleft=TIFFGetStrileByteCount(tif, 0);
 	if (!JPEGFixupTagsSubsamplingSec(&m))
 		TIFFWarningExt(tif->tif_clientdata,module,
 		    "Unable to auto-correct subsampling values, likely corrupt JPEG compressed data in first strip/tile; auto-correcting skipped");
@@ -940,7 +938,10 @@ JPEGFixupTagsSubsamplingReadByte(struct JPEGFixupTagsSubsamplingData* data, uint
 			return(0);
 		if (!data->filepositioned)
 		{
-			TIFFSeekFile(data->tif,data->fileoffset,SEEK_SET);
+			if (TIFFSeekFile(data->tif,data->fileoffset,SEEK_SET) == (toff_t)-1)
+			{
+			    return 0;
+			}
 			data->filepositioned=1;
 		}
 		m=data->buffersize;
@@ -1209,35 +1210,37 @@ JPEGPreDecode(TIFF* tif, uint16 s)
            /* store for all coefficients */
            /* See call to jinit_d_coef_controller() from master_selection() */
            /* in libjpeg */
-            toff_t nRequiredMemory = (toff_t)sp->cinfo.d.image_width *
-                                     sp->cinfo.d.image_height *
-                                     sp->cinfo.d.num_components *
-                                     ((td->td_bitspersample+7)/8);
-            /* BLOCK_SMOOTHING_SUPPORTED is generally defined, so we need */
-            /* to replicate the logic of jinit_d_coef_controller() */
-            if( sp->cinfo.d.progressive_mode )
-                nRequiredMemory *= 3;

-#ifndef TIFF_LIBJPEG_LARGEST_MEM_ALLOC
-#define TIFF_LIBJPEG_LARGEST_MEM_ALLOC (100 * 1024 * 1024)
-#endif
+            /* 1 MB for regular libjpeg usage */
+            toff_t nRequiredMemory = 1024 * 1024;

-            if( nRequiredMemory > TIFF_LIBJPEG_LARGEST_MEM_ALLOC &&
+            for (ci = 0; ci < sp->cinfo.d.num_components; ci++) {
+                const jpeg_component_info *compptr = &(sp->cinfo.d.comp_info[ci]);
+                if( compptr->h_samp_factor > 0 && compptr->v_samp_factor > 0 )
+                {
+                    nRequiredMemory += (toff_t)(
+                        ((compptr->width_in_blocks + compptr->h_samp_factor - 1) / compptr->h_samp_factor)) *
+                        ((compptr->height_in_blocks + compptr->v_samp_factor - 1) / compptr->v_samp_factor) *
+                        sizeof(JBLOCK);
+                }
+            }
+
+            if( sp->cinfo.d.mem->max_memory_to_use > 0 &&
+                nRequiredMemory > (toff_t)(sp->cinfo.d.mem->max_memory_to_use) &&
                getenv("LIBTIFF_ALLOW_LARGE_LIBJPEG_MEM_ALLOC") == NULL )
            {
-                    TIFFErrorExt(tif->tif_clientdata, module,
-                        "Reading this strip would require libjpeg to allocate "
-                        "at least %u bytes. "
-                        "This is disabled since above the %u threshold. "
-                        "You may override this restriction by defining the "
-                        "LIBTIFF_ALLOW_LARGE_LIBJPEG_MEM_ALLOC environment variable, "
-                        "or recompile libtiff by defining the "
-                        "TIFF_LIBJPEG_LARGEST_MEM_ALLOC macro to a value greater "
-                        "than %u",
-                        (unsigned)nRequiredMemory,
-                        (unsigned)TIFF_LIBJPEG_LARGEST_MEM_ALLOC,
-                        (unsigned)TIFF_LIBJPEG_LARGEST_MEM_ALLOC);
-                    return (0);
+                TIFFErrorExt(tif->tif_clientdata, module,
+                    "Reading this image would require libjpeg to allocate "
+                    "at least %u bytes. "
+                    "This is disabled since above the %u threshold. "
+                    "You may override this restriction by defining the "
+                    "LIBTIFF_ALLOW_LARGE_LIBJPEG_MEM_ALLOC environment variable, "
+                    "or setting the JPEGMEM environment variable to a value greater "
+                    "or equal to '%uM'",
+                    (unsigned)(nRequiredMemory),
+                    (unsigned)(sp->cinfo.d.mem->max_memory_to_use),
+                    (unsigned)((nRequiredMemory + 1000000 - 1) / 1000000));
+                return 0;
            }
        }

@@ -1566,7 +1569,7 @@ JPEGDecodeRaw(TIFF* tif, uint8* buf, tmsize_t cc, uint16 s)
 					JSAMPLE *outptr = (JSAMPLE*)tmpbuf + clumpoffset;
 #else
 					JSAMPLE *outptr = (JSAMPLE*)buf + clumpoffset;
-					if (cc < (tmsize_t) (clumpoffset + samples_per_clump*(clumps_per_line-1) + hsamp)) {
+					if (cc < (tmsize_t)(clumpoffset + (tmsize_t)samples_per_clump*(clumps_per_line-1) + hsamp)) {
 						TIFFErrorExt(tif->tif_clientdata, "JPEGDecodeRaw",
 							     "application buffer not large enough for all data, possible subsampling issue");
 						return 0;
@@ -2126,8 +2129,8 @@ JPEGEncodeRaw(TIFF* tif, uint8* buf, tmsize_t cc, uint16 s)
 	/* data is expected to be supplied in multiples of a clumpline */
 	/* a clumpline is equivalent to v_sampling desubsampled scanlines */
 	/* TODO: the following calculation of bytesperclumpline, should substitute calculation of sp->bytesperline, except that it is per v_sampling lines */
-	bytesperclumpline = (((sp->cinfo.c.image_width+sp->h_sampling-1)/sp->h_sampling)
-			     *(sp->h_sampling*sp->v_sampling+2)*sp->cinfo.c.data_precision+7)
+	bytesperclumpline = ((((tmsize_t)sp->cinfo.c.image_width+sp->h_sampling-1)/sp->h_sampling)
+			     *((tmsize_t)sp->h_sampling*sp->v_sampling+2)*sp->cinfo.c.data_precision+7)
 			    /8;

 	nrows = ( cc / bytesperclumpline ) * sp->v_sampling;
@@ -2347,7 +2350,7 @@ JPEGVGetField(TIFF* tif, uint32 tag, va_list ap)
 	switch (tag) {
 		case TIFFTAG_JPEGTABLES:
 			*va_arg(ap, uint32*) = sp->jpegtables_length;
-			*va_arg(ap, void**) = sp->jpegtables;
+			*va_arg(ap, const void**) = sp->jpegtables;
 			break;
 		case TIFFTAG_JPEGQUALITY:
 			*va_arg(ap, int*) = sp->jpegquality;
@@ -2482,6 +2485,7 @@ TIFFInitJPEG(TIFF* tif, int scheme)
 {
 	JPEGState* sp;

+        (void)scheme;
 	assert(scheme == COMPRESSION_JPEG);

 	/*
@@ -193,6 +193,7 @@ LogL16Decode(TIFF* tif, uint8* op, tmsize_t occ, uint16 s)
 	tmsize_t cc;
 	int rc;

+        (void)s;
 	assert(s == 0);
 	assert(sp != NULL);

@@ -266,6 +267,7 @@ LogLuvDecode24(TIFF* tif, uint8* op, tmsize_t occ, uint16 s)
 	unsigned char* bp;
 	uint32* tp;

+        (void)s;
 	assert(s == 0);
 	assert(sp != NULL);

@@ -326,6 +328,7 @@ LogLuvDecode32(TIFF* tif, uint8* op, tmsize_t occ, uint16 s)
 	tmsize_t cc;
 	int rc;

+        (void)s;
 	assert(s == 0);
 	sp = DecoderState(tif);
 	assert(sp != NULL);
@@ -447,6 +450,7 @@ LogL16Encode(TIFF* tif, uint8* bp, tmsize_t cc, uint16 s)
 	int rc=0, mask;
 	tmsize_t beg;

+        (void)s;
 	assert(s == 0);
 	assert(sp != NULL);
 	npixels = cc / sp->pixel_size;
@@ -541,6 +545,7 @@ LogLuvEncode24(TIFF* tif, uint8* bp, tmsize_t cc, uint16 s)
 	uint8* op;
 	uint32* tp;

+        (void)s;
 	assert(s == 0);
 	assert(sp != NULL);
 	npixels = cc / sp->pixel_size;
@@ -598,6 +603,7 @@ LogLuvEncode32(TIFF* tif, uint8* bp, tmsize_t cc, uint16 s)
 	int rc=0, mask;
 	tmsize_t beg;

+        (void)s;
 	assert(s == 0);
 	assert(sp != NULL);

@@ -742,7 +748,7 @@ LogLuvEncodeTile(TIFF* tif, uint8* bp, tmsize_t cc, uint16 s)
 #undef exp2  /* Conflict with C'99 function */
 #define exp2(x)		exp(M_LN2*(x))

-static int itrunc(double x, int m)
+static int tiff_itrunc(double x, int m)
 {
    if( m == SGILOGENCODE_NODITHER )
        return (int)x;
@@ -777,9 +783,9 @@ LogL16fromY(double Y, int em)	/* get 16-bit LogL from Y */
 	if (Y <= -1.8371976e19)
 		return (0xffff);
 	if (Y > 5.4136769e-20)
-		return itrunc(256.*(log2(Y) + 64.), em);
+		return tiff_itrunc(256.*(log2(Y) + 64.), em);
 	if (Y < -5.4136769e-20)
-		return (~0x7fff | itrunc(256.*(log2(-Y) + 64.), em));
+		return (~0x7fff | tiff_itrunc(256.*(log2(-Y) + 64.), em));
 	return (0);
 }

@@ -855,7 +861,7 @@ LogL10fromY(double Y, int em)	/* get 10-bit LogL from Y */
 	else if (Y <= .00024283)
 		return (0);
 	else
-		return itrunc(64.*(log2(Y) + 12.), em);
+		return tiff_itrunc(64.*(log2(Y) + 12.), em);
 }

 #define NANGLES		100
@@ -925,12 +931,12 @@ uv_encode(double u, double v, int em)	/* encode (u',v') coordinates */

 	if (v < UV_VSTART)
 		return oog_encode(u, v);
-	vi = itrunc((v - UV_VSTART)*(1./UV_SQSIZ), em);
+	vi = tiff_itrunc((v - UV_VSTART)*(1./UV_SQSIZ), em);
 	if (vi >= UV_NVS)
 		return oog_encode(u, v);
 	if (u < uv_row[vi].ustart)
 		return oog_encode(u, v);
-	ui = itrunc((u - uv_row[vi].ustart)*(1./UV_SQSIZ), em);
+	ui = tiff_itrunc((u - uv_row[vi].ustart)*(1./UV_SQSIZ), em);
 	if (ui >= uv_row[vi].nus)
 		return oog_encode(u, v);

@@ -1099,7 +1105,7 @@ Luv24fromLuv48(LogLuvState* sp, uint8* op, tmsize_t n)
 		else if (sp->encode_meth == SGILOGENCODE_NODITHER)
 			Le = (luv3[0]-3314) >> 2;
 		else
-			Le = itrunc(.25*(luv3[0]-3314.), sp->encode_meth);
+			Le = tiff_itrunc(.25*(luv3[0]-3314.), sp->encode_meth);

 		Ce = uv_encode((luv3[1]+.5)/(1<<15), (luv3[2]+.5)/(1<<15),
 					sp->encode_meth);
@@ -1155,10 +1161,10 @@ LogLuv32fromXYZ(float XYZ[3], int em)
 		v = 9.*XYZ[1] / s;
 	}
 	if (u <= 0.) ue = 0;
-	else ue = itrunc(UVSCALE*u, em);
+	else ue = tiff_itrunc(UVSCALE*u, em);
 	if (ue > 255) ue = 255;
 	if (v <= 0.) ve = 0;
-	else ve = itrunc(UVSCALE*v, em);
+	else ve = tiff_itrunc(UVSCALE*v, em);
 	if (ve > 255) ve = 255;
 					/* combine encodings */
 	return (Le << 16 | ue << 8 | ve);
@@ -1238,8 +1244,8 @@ Luv32fromLuv48(LogLuvState* sp, uint8* op, tmsize_t n)
 	}
 	while (n-- > 0) {
 		*luv++ = (uint32)luv3[0] << 16 |
-	(itrunc(luv3[1]*(UVSCALE/(1<<15)), sp->encode_meth) << 8 & 0xff00) |
-		(itrunc(luv3[2]*(UVSCALE/(1<<15)), sp->encode_meth) & 0xff);
+	(tiff_itrunc(luv3[1]*(UVSCALE/(1<<15)), sp->encode_meth) << 8 & 0xff00) |
+		(tiff_itrunc(luv3[2]*(UVSCALE/(1<<15)), sp->encode_meth) & 0xff);
 		luv3 += 3;
 	}
 }
@@ -1269,16 +1275,10 @@ LogL16GuessDataFmt(TIFFDirectory *td)
 	return (SGILOGDATAFMT_UNKNOWN);
 }

-
-#define TIFF_SIZE_T_MAX ((size_t) ~ ((size_t)0))
-#define TIFF_TMSIZE_T_MAX (tmsize_t)(TIFF_SIZE_T_MAX >> 1)
-
 static tmsize_t
 multiply_ms(tmsize_t m1, tmsize_t m2)
 {
-        if( m1 == 0 || m2 > TIFF_TMSIZE_T_MAX / m1 )
-            return 0;
-        return m1 * m2;
+        return _TIFFMultiplySSize(NULL, m1, m2, NULL);
 }

 static int
@@ -1512,7 +1512,7 @@ LogLuvSetupEncode(TIFF* tif)
 	switch (td->td_photometric) {
 	case PHOTOMETRIC_LOGLUV:
 		if (!LogLuvInitState(tif))
-			break;
+			return (0);
 		if (td->td_compression == COMPRESSION_SGILOG24) {
 			tif->tif_encoderow = LogLuvEncode24;
 			switch (sp->user_datafmt) {
@@ -1545,7 +1545,7 @@ LogLuvSetupEncode(TIFF* tif)
 		break;
 	case PHOTOMETRIC_LOGL:
 		if (!LogL16InitState(tif))
-			break;
+			return (0);
 		tif->tif_encoderow = LogL16Encode;  
 		switch (sp->user_datafmt) {
 		case SGILOGDATAFMT_FLOAT:
@@ -1561,7 +1561,7 @@ LogLuvSetupEncode(TIFF* tif)
 		TIFFErrorExt(tif->tif_clientdata, module,
 		    "Inappropriate photometric interpretation %d for SGILog compression; %s",
 		    td->td_photometric, "must be either LogLUV or LogL");
-		break;
+		return (0);
 	}
 	sp->encoder_state = 1;
 	return (1);
@@ -300,7 +300,8 @@ LZMAEncode(TIFF* tif, uint8* bp, tmsize_t cc, uint16 s)
 		}
 		if (sp->stream.avail_out == 0) {
 			tif->tif_rawcc = tif->tif_rawdatasize;
-			TIFFFlushData1(tif);
+			if (!TIFFFlushData1(tif))
+				return 0;
 			sp->stream.next_out = tif->tif_rawdata;
 			sp->stream.avail_out = (size_t)tif->tif_rawdatasize;  /* this is a safe typecast, as check is made already in LZMAPreEncode */
 		}
@@ -328,7 +329,8 @@ LZMAPostEncode(TIFF* tif)
 			if ((tmsize_t)sp->stream.avail_out != tif->tif_rawdatasize) {
 				tif->tif_rawcc =
 					tif->tif_rawdatasize - sp->stream.avail_out;
-				TIFFFlushData1(tif);
+				if (!TIFFFlushData1(tif))
+					return 0;
 				sp->stream.next_out = tif->tif_rawdata;
 				sp->stream.avail_out = (size_t)tif->tif_rawdatasize;  /* this is a safe typecast, as check is made already in ZIPPreEncode */
 			}
@@ -418,6 +420,7 @@ TIFFInitLZMA(TIFF* tif, int scheme)
 	LZMAState* sp;
 	lzma_stream tmp_stream = LZMA_STREAM_INIT;

+        (void)scheme;
 	assert( scheme == COMPRESSION_LZMA );

 	/*
@@ -214,19 +214,16 @@ LZWSetupDecode(TIFF* tif)
 			return (0);
 		}

-		DecoderState(tif)->dec_codetab = NULL;
-		DecoderState(tif)->dec_decode = NULL;
+		sp = DecoderState(tif);
+		sp->dec_codetab = NULL;
+		sp->dec_decode = NULL;

 		/*
 		 * Setup predictor setup.
 		 */
 		(void) TIFFPredictorInit(tif);
-
-		sp = DecoderState(tif);
 	}

-	assert(sp != NULL);
-
 	if (sp->dec_codetab == NULL) {
 		sp->dec_codetab = (code_t*)_TIFFmalloc(CSIZE*sizeof (code_t));
 		if (sp->dec_codetab == NULL) {
@@ -1161,6 +1158,7 @@ int
 TIFFInitLZW(TIFF* tif, int scheme)
 {
 	static const char module[] = "TIFFInitLZW";
+        (void)scheme;
 	assert(scheme == COMPRESSION_LZW);
 	/*
 	 * Allocate state block so tag methods have storage to record values.
@@ -1218,7 +1216,7 @@ bad:
 * from this software without specific prior written permission.
 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR
 * IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED
- * WARRANTIES OF MERCHANTIBILITY AND FITNESS FOR A PARTICULAR PURPOSE.
+ * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
 */
 #endif /* LZW_SUPPORT */

@@ -74,7 +74,7 @@
   or errors, up to the point where either these values are read, or it's clear they
   aren't there. This means that some of the data is read twice, but we feel speed
   in correcting these values is important enough to warrant this sacrifice. Although
-   there is currently no define or other configuration mechanism to disable this behaviour,
+   there is currently no define or other configuration mechanism to disable this behavior,
   the actual header scanning is build to robustly respond with error report if it
   should encounter an uncorrected mismatch of subsampling values. See
   OJPEGReadHeaderInfoSecStreamSof.
@@ -243,6 +243,7 @@ typedef enum {
 typedef struct {
 	TIFF* tif;
        int decoder_ok;
+        int error_in_raw_data_decoding;
 	#ifndef LIBJPEG_ENCAP_EXTERNAL
 	JMP_BUF exit_jmpbuf;
 	#endif
@@ -420,6 +421,7 @@ TIFFInitOJPEG(TIFF* tif, int scheme)
 	static const char module[]="TIFFInitOJPEG";
 	OJPEGState* sp;

+        (void)scheme;
 	assert(scheme==COMPRESSION_OJPEG);

        /*
@@ -497,15 +499,15 @@ OJPEGVGetField(TIFF* tif, uint32 tag, va_list ap)
 			break;
 		case TIFFTAG_JPEGQTABLES:
 			*va_arg(ap,uint32*)=(uint32)sp->qtable_offset_count;
-			*va_arg(ap,void**)=(void*)sp->qtable_offset; 
+			*va_arg(ap,const void**)=(const void*)sp->qtable_offset;
 			break;
 		case TIFFTAG_JPEGDCTABLES:
 			*va_arg(ap,uint32*)=(uint32)sp->dctable_offset_count;
-			*va_arg(ap,void**)=(void*)sp->dctable_offset;  
+			*va_arg(ap,const void**)=(const void*)sp->dctable_offset;
 			break;
 		case TIFFTAG_JPEGACTABLES:
 			*va_arg(ap,uint32*)=(uint32)sp->actable_offset_count;
-			*va_arg(ap,void**)=(void*)sp->actable_offset;
+			*va_arg(ap,const void**)=(const void*)sp->actable_offset;
 			break;
 		case TIFFTAG_JPEGPROC:
 			*va_arg(ap,uint16*)=(uint16)sp->jpeg_proc;
@@ -657,7 +659,7 @@ static int
 OJPEGSetupDecode(TIFF* tif)
 {
 	static const char module[]="OJPEGSetupDecode";
-	TIFFWarningExt(tif->tif_clientdata,module,"Depreciated and troublesome old-style JPEG compression mode, please convert to new-style JPEG compression and notify vendor of writing software");
+	TIFFWarningExt(tif->tif_clientdata,module,"Deprecated and troublesome old-style JPEG compression mode, please convert to new-style JPEG compression and notify vendor of writing software");
 	return(1);
 }

@@ -678,7 +680,7 @@ OJPEGPreDecode(TIFF* tif, uint16 s)
 		if (OJPEGReadSecondarySos(tif,s)==0)
 			return(0);
 	}
-	if isTiled(tif)
+	if (isTiled(tif))
 		m=tif->tif_curtile;
 	else
 		m=tif->tif_curstrip;
@@ -742,6 +744,7 @@ OJPEGPreDecodeSkipRaw(TIFF* tif)
 		}
 		m-=sp->subsampling_convert_clines-sp->subsampling_convert_state;
 		sp->subsampling_convert_state=0;
+                sp->error_in_raw_data_decoding=0;
 	}
 	while (m>=sp->subsampling_convert_clines)
 	{
@@ -792,6 +795,10 @@ OJPEGDecode(TIFF* tif, uint8* buf, tmsize_t cc, uint16 s)
            TIFFErrorExt(tif->tif_clientdata,module,"Cannot decode: decoder not correctly initialized");
            return 0;
        }
+        if( sp->error_in_raw_data_decoding )
+        {
+            return 0;
+        }
 	if (sp->libjpeg_jpeg_query_style==0)
 	{
 		if (OJPEGDecodeRaw(tif,buf,cc)==0)
@@ -832,7 +839,10 @@ OJPEGDecodeRaw(TIFF* tif, uint8* buf, tmsize_t cc)
 		if (sp->subsampling_convert_state==0)
 		{
 			if (jpeg_read_raw_data_encap(sp,&(sp->libjpeg_jpeg_decompress_struct),sp->subsampling_convert_ycbcrimage,sp->subsampling_ver*8)==0)
+			{
+				sp->error_in_raw_data_decoding = 1;
 				return(0);
+			}
 		}
 		oy=sp->subsampling_convert_ybuf+sp->subsampling_convert_state*sp->subsampling_ver*sp->subsampling_convert_ylinelen;
 		ocb=sp->subsampling_convert_cbbuf+sp->subsampling_convert_state*sp->subsampling_convert_clinelen;
@@ -990,7 +1000,6 @@ OJPEGSubsamplingCorrect(TIFF* tif)
 	OJPEGState* sp=(OJPEGState*)tif->tif_data;
 	uint8 mh;
 	uint8 mv;
-        _TIFFFillStriles( tif );
        
 	assert(sp->subsamplingcorrect_done==0);
 	if ((tif->tif_dir.td_samplesperpixel!=3) || ((tif->tif_dir.td_photometric!=PHOTOMETRIC_YCBCR) &&
@@ -1046,7 +1055,7 @@ OJPEGReadHeaderInfo(TIFF* tif)
 	assert(sp->readheader_done==0);
 	sp->image_width=tif->tif_dir.td_imagewidth;
 	sp->image_length=tif->tif_dir.td_imagelength;
-	if isTiled(tif)
+	if (isTiled(tif))
 	{
 		sp->strile_width=tif->tif_dir.td_tilewidth;
 		sp->strile_length=tif->tif_dir.td_tilelength;
@@ -1056,6 +1065,8 @@ OJPEGReadHeaderInfo(TIFF* tif)
 	{
 		sp->strile_width=sp->image_width;
 		sp->strile_length=tif->tif_dir.td_rowsperstrip;
+                if( sp->strile_length == (uint32)-1 )
+                    sp->strile_length = sp->image_length;
 		sp->strile_length_total=sp->image_length;
 	}
 	if (tif->tif_dir.td_samplesperpixel==1)
@@ -1082,6 +1093,12 @@ OJPEGReadHeaderInfo(TIFF* tif)
 	}
 	if (sp->strile_length<sp->image_length)
 	{
+		if (((sp->subsampling_hor!=1) && (sp->subsampling_hor!=2) && (sp->subsampling_hor!=4)) ||
+		    ((sp->subsampling_ver!=1) && (sp->subsampling_ver!=2) && (sp->subsampling_ver!=4)))
+		{
+			TIFFErrorExt(tif->tif_clientdata,module,"Invalid subsampling values");
+			return(0);
+		}
 		if (sp->strile_length%(sp->subsampling_ver*8)!=0)
 		{
 			TIFFErrorExt(tif->tif_clientdata,module,"Incompatible vertical subsampling and image strip/tile length");
@@ -1197,7 +1214,13 @@ OJPEGWriteHeaderInfo(TIFF* tif)
 			sp->subsampling_convert_ybuflen=sp->subsampling_convert_ylinelen*sp->subsampling_convert_ylines;
 			sp->subsampling_convert_cbuflen=sp->subsampling_convert_clinelen*sp->subsampling_convert_clines;
 			sp->subsampling_convert_ycbcrbuflen=sp->subsampling_convert_ybuflen+2*sp->subsampling_convert_cbuflen;
-			sp->subsampling_convert_ycbcrbuf=_TIFFmalloc(sp->subsampling_convert_ycbcrbuflen);
+                        /* The calloc is not normally necessary, except in some edge/broken cases */
+                        /* for example for a tiled image of height 1 with a tile height of 1 and subsampling_hor=subsampling_ver=2 */
+                        /* In that case, libjpeg will only fill the 8 first lines of the 16 lines */
+                        /* See https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=16844 */
+                        /* Even if this case is allowed (?), its handling is broken because OJPEGPreDecode() should also likely */
+                        /* reset subsampling_convert_state to 0 when changing tile. */
+			sp->subsampling_convert_ycbcrbuf=_TIFFcalloc(1, sp->subsampling_convert_ycbcrbuflen);
 			if (sp->subsampling_convert_ycbcrbuf==0)
 			{
 				TIFFErrorExt(tif->tif_clientdata,module,"Out of memory");
@@ -1223,10 +1246,11 @@ OJPEGWriteHeaderInfo(TIFF* tif)
 				*m++=sp->subsampling_convert_cbbuf+n*sp->subsampling_convert_clinelen;
 			for (n=0; n<sp->subsampling_convert_clines; n++)
 				*m++=sp->subsampling_convert_crbuf+n*sp->subsampling_convert_clinelen;
-			sp->subsampling_convert_clinelenout=((sp->strile_width+sp->subsampling_hor-1)/sp->subsampling_hor);
+			sp->subsampling_convert_clinelenout=sp->strile_width/sp->subsampling_hor + ((sp->strile_width % sp->subsampling_hor) != 0 ? 1 : 0);
 			sp->subsampling_convert_state=0;
+			sp->error_in_raw_data_decoding=0;
 			sp->bytes_per_line=sp->subsampling_convert_clinelenout*(sp->subsampling_ver*sp->subsampling_hor+2);
-			sp->lines_per_strile=((sp->strile_length+sp->subsampling_ver-1)/sp->subsampling_ver);
+			sp->lines_per_strile=sp->strile_length/sp->subsampling_ver + ((sp->strile_length % sp->subsampling_ver) != 0 ? 1 : 0);
 			sp->subsampling_convert_log=1;
 		}
 	}
@@ -1240,6 +1264,26 @@ OJPEGWriteHeaderInfo(TIFF* tif)
 	}
 	if (jpeg_start_decompress_encap(sp,&(sp->libjpeg_jpeg_decompress_struct))==0)
 		return(0);
+        if(sp->libjpeg_jpeg_decompress_struct.image_width != sp->strile_width ) {
+            TIFFErrorExt(tif->tif_clientdata,module,
+                         "jpeg_start_decompress() returned image_width = %d, "
+                         "expected %d",
+                         sp->libjpeg_jpeg_decompress_struct.image_width,
+                         sp->strile_width);
+            return 0;
+        }
+        if(sp->libjpeg_jpeg_decompress_struct.max_h_samp_factor != sp->subsampling_hor ||
+           sp->libjpeg_jpeg_decompress_struct.max_v_samp_factor != sp->subsampling_ver) {
+            TIFFErrorExt(tif->tif_clientdata,module,
+                         "jpeg_start_decompress() returned max_h_samp_factor = %d "
+                         "and max_v_samp_factor = %d, expected %d and %d",
+                         sp->libjpeg_jpeg_decompress_struct.max_h_samp_factor,
+                         sp->libjpeg_jpeg_decompress_struct.max_v_samp_factor,
+                         sp->subsampling_hor,
+                         sp->subsampling_ver);
+            return 0;
+        }
+
 	sp->writeheader_done=1;
 	return(1);
 }
@@ -1272,7 +1316,9 @@ OJPEGReadHeaderInfoSec(TIFF* tif)
 		}
 		else
 		{
-			if ((sp->jpeg_interchange_format_length==0) || (sp->jpeg_interchange_format+sp->jpeg_interchange_format_length>sp->file_size))
+			if ((sp->jpeg_interchange_format_length==0) ||
+                            (sp->jpeg_interchange_format > TIFF_UINT64_MAX - sp->jpeg_interchange_format_length) ||
+                            (sp->jpeg_interchange_format+sp->jpeg_interchange_format_length>sp->file_size))
 				sp->jpeg_interchange_format_length=sp->file_size-sp->jpeg_interchange_format;
 		}
 	}
@@ -1989,32 +2035,30 @@ OJPEGReadBufferFill(OJPEGState* sp)
 				sp->in_buffer_source=osibsStrile;
                                break;
 			case osibsStrile:
-				if (!_TIFFFillStriles( sp->tif ) 
-				    || sp->tif->tif_dir.td_stripoffset == NULL
-				    || sp->tif->tif_dir.td_stripbytecount == NULL)
-					return 0;
-
 				if (sp->in_buffer_next_strile==sp->in_buffer_strile_count)
 					sp->in_buffer_source=osibsEof;
 				else
 				{
-					sp->in_buffer_file_pos=sp->tif->tif_dir.td_stripoffset[sp->in_buffer_next_strile];
+					int err = 0;
+					sp->in_buffer_file_pos=TIFFGetStrileOffsetWithErr(sp->tif, sp->in_buffer_next_strile, &err);
+					if( err )
+						return 0;
 					if (sp->in_buffer_file_pos!=0)
 					{
+						uint64 bytecount = TIFFGetStrileByteCountWithErr(sp->tif, sp->in_buffer_next_strile, &err);
+						if( err )
+							return 0;
 						if (sp->in_buffer_file_pos>=sp->file_size)
 							sp->in_buffer_file_pos=0;
-						else if (sp->tif->tif_dir.td_stripbytecount==NULL)
+						else if (bytecount==0)
 							sp->in_buffer_file_togo=sp->file_size-sp->in_buffer_file_pos;
 						else
 						{
-							if (sp->tif->tif_dir.td_stripbytecount == 0) {
-								TIFFErrorExt(sp->tif->tif_clientdata,sp->tif->tif_name,"Strip byte counts are missing");
-								return(0);
-							}
-							sp->in_buffer_file_togo=sp->tif->tif_dir.td_stripbytecount[sp->in_buffer_next_strile];
+							sp->in_buffer_file_togo=bytecount;
 							if (sp->in_buffer_file_togo==0)
 								sp->in_buffer_file_pos=0;
-							else if (sp->in_buffer_file_pos+sp->in_buffer_file_togo>sp->file_size)
+							else if (sp->in_buffer_file_pos > TIFF_UINT64_MAX - sp->in_buffer_file_togo || 
+                                                                sp->in_buffer_file_pos+sp->in_buffer_file_togo>sp->file_size)
 								sp->in_buffer_file_togo=sp->file_size-sp->in_buffer_file_pos;
 						}
 					}
@@ -104,6 +104,7 @@ TIFFClientOpen(
 		} n;
 		n.a8[0]=1;
 		n.a8[1]=0;
+                (void)n;
 		#ifdef WORDS_BIGENDIAN
 		assert(n.a16==256);
 		#else
@@ -131,6 +132,7 @@ TIFFClientOpen(
 	if (!readproc || !writeproc || !seekproc || !closeproc || !sizeproc) {
 		TIFFErrorExt(clientdata, module,
 		    "One of the client procedures is NULL pointer.");
+		_TIFFfree(tif);
 		goto bad2;
 	}
 	tif->tif_readproc = readproc;
@@ -164,7 +166,7 @@ TIFFClientOpen(
 	/*
 	 * Process library-specific flags in the open mode string.
 	 * The following flags may be used to control intrinsic library
-	 * behaviour that may or may not be desirable (usually for
+	 * behavior that may or may not be desirable (usually for
 	 * compatibility with some application that claims to support
 	 * TIFF but only supports some brain dead idea of what the
 	 * vendor thinks TIFF is):
@@ -181,6 +183,8 @@ TIFFClientOpen(
 	 * 'h' read TIFF header only, do not load the first IFD
 	 * '4' ClassicTIFF for creating a file (default)
 	 * '8' BigTIFF for creating a file
+         * 'D' enable use of deferred strip/tile offset/bytecount array loading.
+         * 'O' on-demand loading of values instead of whole array loading (implies D)
 	 *
 	 * The use of the 'l' and 'b' flags is strongly discouraged.
 	 * These flags are provided solely because numerous vendors,
@@ -203,7 +207,7 @@ TIFFClientOpen(
 	 * not do right now.
 	 *
 	 * The 'M' and 'm' flags are provided because some virtual memory
-	 * systems exhibit poor behaviour when large images are mapped.
+	 * systems exhibit poor behavior when large images are mapped.
 	 * These options permit clients to control the use of memory-mapped
 	 * files on a per-file basis.
 	 *
@@ -262,7 +266,22 @@ TIFFClientOpen(
 				if (m&O_CREAT)
 					tif->tif_flags |= TIFF_BIGTIFF;
 				break;
+			case 'D':
+			        tif->tif_flags |= TIFF_DEFERSTRILELOAD;
+				break;
+			case 'O':
+				if( m == O_RDONLY )
+					tif->tif_flags |= (TIFF_LAZYSTRILELOAD | TIFF_DEFERSTRILELOAD);
+				break;
 		}
+
+#ifdef DEFER_STRILE_LOAD
+        /* Compatibility with old DEFER_STRILE_LOAD compilation flag */
+        /* Probably unneeded, since to the best of my knowledge (E. Rouault) */
+        /* GDAL was the only user of this, and will now use the new 'D' flag */
+        tif->tif_flags |= TIFF_DEFERSTRILELOAD;
+#endif
+
 	/*
 	 * Read in TIFF header.
 	 */
@@ -634,16 +634,10 @@ PixarLogGuessDataFmt(TIFFDirectory *td)
 	return guess;
 }

-#define TIFF_SIZE_T_MAX ((size_t) ~ ((size_t)0))
-#define TIFF_TMSIZE_T_MAX (tmsize_t)(TIFF_SIZE_T_MAX >> 1)
-
 static tmsize_t
 multiply_ms(tmsize_t m1, tmsize_t m2)
 {
-        assert(m1 >= 0 && m2 >= 0);
-        if( m1 == 0 || m2 > TIFF_TMSIZE_T_MAX / m1 )
-            return 0;
-        return m1 * m2;
+        return _TIFFMultiplySSize(NULL, m1, m2, NULL);
 }

 static tmsize_t
@@ -1153,7 +1147,7 @@ PixarLogEncode(TIFF* tif, uint8* bp, tmsize_t cc, uint16 s)

 	llen = sp->stride * td->td_imagewidth;
    /* Check against the number of elements (of size uint16) of sp->tbuf */
-    if( n > (tmsize_t)(td->td_rowsperstrip * llen) )
+    if( n > ((tmsize_t)td->td_rowsperstrip * llen) )
    {
        TIFFErrorExt(tif->tif_clientdata, module,
                     "Too many input bytes provided");
@@ -1206,7 +1200,8 @@ PixarLogEncode(TIFF* tif, uint8* bp, tmsize_t cc, uint16 s)
 		}
 		if (sp->stream.avail_out == 0) {
 			tif->tif_rawcc = tif->tif_rawdatasize;
-			TIFFFlushData1(tif);
+			if (!TIFFFlushData1(tif))
+				return 0;
 			sp->stream.next_out = tif->tif_rawdata;
 			sp->stream.avail_out = (uInt) tif->tif_rawdatasize;  /* this is a safe typecast, as check is made already in PixarLogPreEncode */
 		}
@@ -1236,7 +1231,8 @@ PixarLogPostEncode(TIFF* tif)
 		    if ((tmsize_t)sp->stream.avail_out != tif->tif_rawdatasize) {
 			    tif->tif_rawcc =
 				tif->tif_rawdatasize - sp->stream.avail_out;
-			    TIFFFlushData1(tif);
+			    if (!TIFFFlushData1(tif))
+                                return 0;
 			    sp->stream.next_out = tif->tif_rawdata;
 			    sp->stream.avail_out = (uInt) tif->tif_rawdatasize;  /* this is a safe typecast, as check is made already in PixarLogPreEncode */
 		    }
@@ -1404,6 +1400,7 @@ TIFFInitPixarLog(TIFF* tif, int scheme)

 	PixarLogState* sp;

+        (void)scheme;
 	assert(scheme == COMPRESSION_PIXARLOG);

 	/*
@@ -116,7 +116,7 @@ PredictorSetupDecode(TIFF* tif)
 	TIFFDirectory* td = &tif->tif_dir;

 	/* Note: when PredictorSetup() fails, the effets of setupdecode() */
-	/* will not be "cancelled" so setupdecode() might be robust to */
+	/* will not be "canceled" so setupdecode() might be robust to */
 	/* be called several times. */
 	if (!(*sp->setupdecode)(tif) || !PredictorSetup(tif))
 		return 0;
@@ -270,8 +270,8 @@ PredictorSetupEncode(TIFF* tif)
    }

 /* Remarks related to C standard compliance in all below functions : */
-/* - to avoid any undefined behaviour, we only operate on unsigned types */
-/*   since the behaviour of "overflows" is defined (wrap over) */
+/* - to avoid any undefined behavior, we only operate on unsigned types */
+/*   since the behavior of "overflows" is defined (wrap over) */
 /* - when storing into the byte stream, we explicitly mask with 0xff so */
 /*   as to make icc -check=conversions happy (not necessary by the standard) */

@@ -652,8 +652,6 @@ TIFFPrintDirectory(TIFF* tif, FILE* fd, long flags)
 	if (tif->tif_tagmethods.printdir)
 		(*tif->tif_tagmethods.printdir)(tif, fd, flags);

-        _TIFFFillStriles( tif );
-        
 	if ((flags & TIFFPRINT_STRIPS) &&
 	    TIFFFieldSet(tif,FIELD_STRIPOFFSETS)) {
 		uint32 s;
@@ -665,13 +663,13 @@ TIFFPrintDirectory(TIFF* tif, FILE* fd, long flags)
 #if defined(__WIN32__) && (defined(_MSC_VER) || defined(__MINGW32__))
 			fprintf(fd, "    %3lu: [%8I64u, %8I64u]\n",
 			    (unsigned long) s,
-			    td->td_stripoffset ? (unsigned __int64) td->td_stripoffset[s] : 0,
-			    td->td_stripbytecount ? (unsigned __int64) td->td_stripbytecount[s] : 0);
+			    (unsigned __int64) TIFFGetStrileOffset(tif, s),
+			    (unsigned __int64) TIFFGetStrileByteCount(tif, s));
 #else
 			fprintf(fd, "    %3lu: [%8llu, %8llu]\n",
 			    (unsigned long) s,
-			    td->td_stripoffset ? (unsigned long long) td->td_stripoffset[s] : 0,
-			    td->td_stripbytecount ? (unsigned long long) td->td_stripbytecount[s] : 0);
+			    (unsigned long long) TIFFGetStrileOffset(tif, s),
+			    (unsigned long long) TIFFGetStrileByteCount(tif, s));
 #endif
 	}
 }
@@ -29,9 +29,6 @@
 #include "tiffiop.h"
 #include <stdio.h>

-#define TIFF_SIZE_T_MAX ((size_t) ~ ((size_t)0))
-#define TIFF_TMSIZE_T_MAX (tmsize_t)(TIFF_SIZE_T_MAX >> 1)
-
 int TIFFFillStrip(TIFF* tif, uint32 strip);
 int TIFFFillTile(TIFF* tif, uint32 tile);
 static int TIFFStartStrip(TIFF* tif, uint32 strip);
@@ -49,6 +46,8 @@ TIFFReadRawTile1(TIFF* tif, uint32 tile, void* buf, tmsize_t size, const char* m
 #define THRESHOLD_MULTIPLIER 10
 #define MAX_THRESHOLD (THRESHOLD_MULTIPLIER * THRESHOLD_MULTIPLIER * THRESHOLD_MULTIPLIER * INITIAL_THRESHOLD)

+#define TIFF_INT64_MAX ((((int64)0x7FFFFFFF) << 32) | 0xFFFFFFFF)
+
 /* Read 'size' bytes in tif_rawdata buffer starting at offset 'rawdata_offset'
 * Returns 1 in case of success, 0 otherwise. */
 static int TIFFReadAndRealloc( TIFF* tif, tmsize_t size,
@@ -61,6 +60,22 @@ static int TIFFReadAndRealloc( TIFF* tif, tmsize_t size,
 #endif
        tmsize_t already_read = 0;

+
+#if SIZEOF_SIZE_T != 8
+        /* On 32 bit processes, if the request is large enough, check against */
+        /* file size */
+        if( size > 1000 * 1000 * 1000 )
+        {
+            uint64 filesize = TIFFGetFileSize(tif);
+            if( (uint64)size >= filesize )
+            {
+                TIFFErrorExt(tif->tif_clientdata, module,
+                             "Chunk size requested is larger than file size.");
+                return 0;
+            }
+        }
+#endif
+
        /* On 64 bit processes, read first a maximum of 1 MB, then 10 MB, etc */
        /* so as to avoid allocating too much memory in case the file is too */
        /* short. We could ask for the file size, but this might be */
@@ -175,17 +190,14 @@ TIFFFillStripPartial( TIFF *tif, int strip, tmsize_t read_ahead, int restart )
        tmsize_t to_read;
        tmsize_t read_ahead_mod;
        /* tmsize_t bytecountm; */
-        
-        if (!_TIFFFillStriles( tif ) || !tif->tif_dir.td_stripbytecount)
-            return 0;
-        
+
        /*
         * Expand raw data buffer, if needed, to hold data
         * strip coming from file (perhaps should set upper
         * bound on the size of a buffer we'll use?).
         */

-        /* bytecountm=(tmsize_t) td->td_stripbytecount[strip]; */
+        /* bytecountm=(tmsize_t) TIFFGetStrileByteCount(tif, strip); */

        /* Not completely sure where the * 2 comes from, but probably for */
        /* an exponentional growth strategy of tif_rawdatasize */
@@ -229,7 +241,7 @@ TIFFFillStripPartial( TIFF *tif, int strip, tmsize_t read_ahead, int restart )
        /*
        ** Seek to the point in the file where more data should be read.
        */
-        read_offset = td->td_stripoffset[strip]
+        read_offset = TIFFGetStrileOffset(tif, strip)
                + tif->tif_rawdataoff + tif->tif_rawdataloaded;

        if (!SeekOK(tif, read_offset)) {
@@ -246,10 +258,10 @@ TIFFFillStripPartial( TIFF *tif, int strip, tmsize_t read_ahead, int restart )
                to_read = read_ahead_mod - unused_data;
        else
                to_read = tif->tif_rawdatasize - unused_data;
-        if( (uint64) to_read > td->td_stripbytecount[strip] 
+        if( (uint64) to_read > TIFFGetStrileByteCount(tif, strip)
            - tif->tif_rawdataoff - tif->tif_rawdataloaded )
        {
-                to_read = (tmsize_t) td->td_stripbytecount[strip]
+                to_read = (tmsize_t) TIFFGetStrileByteCount(tif, strip)
                        - tif->tif_rawdataoff - tif->tif_rawdataloaded;
        }

@@ -288,7 +300,7 @@ TIFFFillStripPartial( TIFF *tif, int strip, tmsize_t read_ahead, int restart )
            /* For JPEG, if there are multiple scans (can generally be known */
            /* with the  read_ahead used), we need to read the whole strip */
            if( tif->tif_dir.td_compression==COMPRESSION_JPEG &&
-                (uint64)tif->tif_rawcc < td->td_stripbytecount[strip] )
+                (uint64)tif->tif_rawcc < TIFFGetStrileByteCount(tif, strip) )
            {
                if( TIFFJPEGIsFullStripRequired(tif) )
                {
@@ -347,9 +359,7 @@ TIFFSeek(TIFF* tif, uint32 row, uint16 sample )
         * read it a few lines at a time?
         */
 #if defined(CHUNKY_STRIP_READ_SUPPORT)
-        if (!_TIFFFillStriles( tif ) || !tif->tif_dir.td_stripbytecount)
-            return 0;
-        whole_strip = tif->tif_dir.td_stripbytecount[strip] < 10
+        whole_strip = TIFFGetStrileByteCount(tif, strip) < 10
                || isMapped(tif);
        if( td->td_compression == COMPRESSION_LERC ||
            td->td_compression == COMPRESSION_JBIG )
@@ -402,7 +412,7 @@ TIFFSeek(TIFF* tif, uint32 row, uint16 sample )
        else if( !whole_strip )
        {
                if( ((tif->tif_rawdata + tif->tif_rawdataloaded) - tif->tif_rawcp) < read_ahead 
-                    && (uint64) tif->tif_rawdataoff+tif->tif_rawdataloaded < td->td_stripbytecount[strip] )
+                    && (uint64) tif->tif_rawdataoff+tif->tif_rawdataloaded < TIFFGetStrileByteCount(tif, strip) )
                {
                        if( !TIFFFillStripPartial(tif,strip,read_ahead,0) )
                                return 0;
@@ -599,16 +609,11 @@ static tmsize_t
 TIFFReadRawStrip1(TIFF* tif, uint32 strip, void* buf, tmsize_t size,
    const char* module)
 {
-	TIFFDirectory *td = &tif->tif_dir;
-
-    if (!_TIFFFillStriles( tif ))
-        return ((tmsize_t)(-1));
-        
 	assert((tif->tif_flags&TIFF_NOREADRAW)==0);
 	if (!isMapped(tif)) {
 		tmsize_t cc;

-		if (!SeekOK(tif, td->td_stripoffset[strip])) {
+		if (!SeekOK(tif, TIFFGetStrileOffset(tif, strip))) {
 			TIFFErrorExt(tif->tif_clientdata, module,
 			    "Seek error at scanline %lu, strip %lu",
 			    (unsigned long) tif->tif_row, (unsigned long) strip);
@@ -634,8 +639,8 @@ TIFFReadRawStrip1(TIFF* tif, uint32 strip, void* buf, tmsize_t size,
 	} else {
 		tmsize_t ma = 0;
 		tmsize_t n;
-		if ((td->td_stripoffset[strip] > (uint64)TIFF_TMSIZE_T_MAX)||
-                    ((ma=(tmsize_t)td->td_stripoffset[strip])>tif->tif_size))
+		if ((TIFFGetStrileOffset(tif, strip) > (uint64)TIFF_TMSIZE_T_MAX)||
+                    ((ma=(tmsize_t)TIFFGetStrileOffset(tif, strip))>tif->tif_size))
                {
                    n=0;
                }
@@ -679,12 +684,10 @@ static tmsize_t
 TIFFReadRawStripOrTile2(TIFF* tif, uint32 strip_or_tile, int is_strip,
                        tmsize_t size, const char* module)
 {
-        TIFFDirectory *td = &tif->tif_dir;
-
        assert( !isMapped(tif) );
        assert((tif->tif_flags&TIFF_NOREADRAW)==0);

-        if (!SeekOK(tif, td->td_stripoffset[strip_or_tile])) {
+        if (!SeekOK(tif, TIFFGetStrileOffset(tif, strip_or_tile))) {
            if( is_strip )
            {
                TIFFErrorExt(tif->tif_clientdata, module,
@@ -720,7 +723,7 @@ TIFFReadRawStrip(TIFF* tif, uint32 strip, void* buf, tmsize_t size)
 {
 	static const char module[] = "TIFFReadRawStrip";
 	TIFFDirectory *td = &tif->tif_dir;
-	uint64 bytecount;
+	uint64 bytecount64;
 	tmsize_t bytecountm;

 	if (!TIFFCheckRead(tif, 0))
@@ -738,31 +741,23 @@ TIFFReadRawStrip(TIFF* tif, uint32 strip, void* buf, tmsize_t size)
 		    "Compression scheme does not support access to raw uncompressed data");
 		return ((tmsize_t)(-1));
 	}
-	bytecount = td->td_stripbytecount[strip];
-	if ((int64)bytecount <= 0) {
-#if defined(__WIN32__) && (defined(_MSC_VER) || defined(__MINGW32__))
-		TIFFErrorExt(tif->tif_clientdata, module,
-			     "%I64u: Invalid strip byte count, strip %lu",
-			     (unsigned __int64) bytecount,
-			     (unsigned long) strip);
-#else
-		TIFFErrorExt(tif->tif_clientdata, module,
-			     "%llu: Invalid strip byte count, strip %lu",
-			     (unsigned long long) bytecount,
-			     (unsigned long) strip);
-#endif
-		return ((tmsize_t)(-1));
-	}
-	bytecountm = (tmsize_t)bytecount;
-	if ((uint64)bytecountm!=bytecount) {
-		TIFFErrorExt(tif->tif_clientdata, module, "Integer overflow");
-		return ((tmsize_t)(-1));
-	}
-	if (size != (tmsize_t)(-1) && size < bytecountm)
+	bytecount64 = TIFFGetStrileByteCount(tif, strip);
+	if (size != (tmsize_t)(-1) && (uint64)size <= bytecount64)
 		bytecountm = size;
+	else
+		bytecountm = _TIFFCastUInt64ToSSize(tif, bytecount64, module);
+	if( bytecountm == 0 ) {
+		return ((tmsize_t)(-1));
+	}
 	return (TIFFReadRawStrip1(tif, strip, buf, bytecountm, module));
 }

+TIFF_NOSANITIZE_UNSIGNED_INT_OVERFLOW
+static uint64 NoSanitizeSubUInt64(uint64 a, uint64 b)
+{
+    return a - b;
+}
+
 /*
 * Read the specified strip and setup for decoding. The data buffer is
 * expanded, as necessary, to hold the strip's data.
@@ -773,13 +768,10 @@ TIFFFillStrip(TIFF* tif, uint32 strip)
 	static const char module[] = "TIFFFillStrip";
 	TIFFDirectory *td = &tif->tif_dir;

-        if (!_TIFFFillStriles( tif ) || !tif->tif_dir.td_stripbytecount)
-            return 0;
-
 	if ((tif->tif_flags&TIFF_NOREADRAW)==0)
 	{
-		uint64 bytecount = td->td_stripbytecount[strip];
-		if ((int64)bytecount <= 0) {
+		uint64 bytecount = TIFFGetStrileByteCount(tif, strip);
+		if( bytecount == 0 || bytecount > (uint64)TIFF_INT64_MAX ) {
 #if defined(__WIN32__) && (defined(_MSC_VER) || defined(__MINGW32__))
 			TIFFErrorExt(tif->tif_clientdata, module,
 				"Invalid strip byte count %I64u, strip %lu",
@@ -806,7 +798,7 @@ TIFFFillStrip(TIFF* tif, uint32 strip)
 			    (bytecount - 4096) / 10 > (uint64)stripsize  )
 			{
 				uint64 newbytecount = (uint64)stripsize * 10 + 4096;
-				if( (int64)newbytecount >= 0 )
+				if( newbytecount == 0 || newbytecount > (uint64)TIFF_INT64_MAX )
 				{
 #if defined(__WIN32__) && (defined(_MSC_VER) || defined(__MINGW32__))
 					TIFFWarningExt(tif->tif_clientdata, module,
@@ -831,13 +823,13 @@ TIFFFillStrip(TIFF* tif, uint32 strip)
 			 * We must check for overflow, potentially causing
 			 * an OOB read. Instead of simple
 			 *
-			 *  td->td_stripoffset[strip]+bytecount > tif->tif_size
+			 *  TIFFGetStrileOffset(tif, strip)+bytecount > tif->tif_size
 			 *
 			 * comparison (which can overflow) we do the following
 			 * two comparisons:
 			 */
 			if (bytecount > (uint64)tif->tif_size ||
-			    td->td_stripoffset[strip] > (uint64)tif->tif_size - bytecount) {
+			    TIFFGetStrileOffset(tif, strip) > (uint64)tif->tif_size - bytecount) {
 				/*
 				 * This error message might seem strange, but
 				 * it's what would happen if a read were done
@@ -849,7 +841,7 @@ TIFFFillStrip(TIFF* tif, uint32 strip)
 					"Read error on strip %lu; "
 					"got %I64u bytes, expected %I64u",
 					(unsigned long) strip,
-					(unsigned __int64) tif->tif_size - td->td_stripoffset[strip],
+					(unsigned __int64) NoSanitizeSubUInt64(tif->tif_size, TIFFGetStrileOffset(tif, strip)),
 					(unsigned __int64) bytecount);
 #else
 				TIFFErrorExt(tif->tif_clientdata, module,
@@ -857,7 +849,7 @@ TIFFFillStrip(TIFF* tif, uint32 strip)
 					"Read error on strip %lu; "
 					"got %llu bytes, expected %llu",
 					(unsigned long) strip,
-					(unsigned long long) tif->tif_size - td->td_stripoffset[strip],
+					(unsigned long long) NoSanitizeSubUInt64(tif->tif_size, TIFFGetStrileOffset(tif, strip)),
 					(unsigned long long) bytecount);
 #endif
 				tif->tif_curstrip = NOSTRIP;
@@ -886,7 +878,7 @@ TIFFFillStrip(TIFF* tif, uint32 strip)
 			}
 			tif->tif_flags &= ~TIFF_MYBUFFER;
 			tif->tif_rawdatasize = (tmsize_t)bytecount;
-			tif->tif_rawdata = tif->tif_base + (tmsize_t)td->td_stripoffset[strip];
+			tif->tif_rawdata = tif->tif_base + (tmsize_t)TIFFGetStrileOffset(tif, strip);
                        tif->tif_rawdataoff = 0;
                        tif->tif_rawdataloaded = (tmsize_t) bytecount;

@@ -1101,16 +1093,11 @@ _TIFFReadEncodedTileAndAllocBuffer(TIFF* tif, uint32 tile,
 static tmsize_t
 TIFFReadRawTile1(TIFF* tif, uint32 tile, void* buf, tmsize_t size, const char* module)
 {
-	TIFFDirectory *td = &tif->tif_dir;
-
-    if (!_TIFFFillStriles( tif ))
-        return ((tmsize_t)(-1));
-
 	assert((tif->tif_flags&TIFF_NOREADRAW)==0);
 	if (!isMapped(tif)) {
 		tmsize_t cc;

-		if (!SeekOK(tif, td->td_stripoffset[tile])) {
+		if (!SeekOK(tif, TIFFGetStrileOffset(tif, tile))) {
 			TIFFErrorExt(tif->tif_clientdata, module,
 			    "Seek error at row %lu, col %lu, tile %lu",
 			    (unsigned long) tif->tif_row,
@@ -1140,9 +1127,9 @@ TIFFReadRawTile1(TIFF* tif, uint32 tile, void* buf, tmsize_t size, const char* m
 	} else {
 		tmsize_t ma,mb;
 		tmsize_t n;
-		ma=(tmsize_t)td->td_stripoffset[tile];
+		ma=(tmsize_t)TIFFGetStrileOffset(tif, tile);
 		mb=ma+size;
-		if ((td->td_stripoffset[tile] > (uint64)TIFF_TMSIZE_T_MAX)||(ma>tif->tif_size))
+		if ((TIFFGetStrileOffset(tif, tile) > (uint64)TIFF_TMSIZE_T_MAX)||(ma>tif->tif_size))
 			n=0;
 		else if ((mb<ma)||(mb<size)||(mb>tif->tif_size))
 			n=tif->tif_size-ma;
@@ -1198,13 +1185,12 @@ TIFFReadRawTile(TIFF* tif, uint32 tile, void* buf, tmsize_t size)
 		"Compression scheme does not support access to raw uncompressed data");
 		return ((tmsize_t)(-1));
 	}
-	bytecount64 = td->td_stripbytecount[tile];
-	if (size != (tmsize_t)(-1) && (uint64)size < bytecount64)
-		bytecount64 = (uint64)size;
-	bytecountm = (tmsize_t)bytecount64;
-	if ((uint64)bytecountm!=bytecount64)
-	{
-		TIFFErrorExt(tif->tif_clientdata,module,"Integer overflow");
+	bytecount64 = TIFFGetStrileByteCount(tif, tile);
+	if (size != (tmsize_t)(-1) && (uint64)size <= bytecount64)
+		bytecountm = size;
+	else
+		bytecountm = _TIFFCastUInt64ToSSize(tif, bytecount64, module);
+	if( bytecountm == 0 ) {
 		return ((tmsize_t)(-1));
 	}
 	return (TIFFReadRawTile1(tif, tile, buf, bytecountm, module));
@@ -1220,13 +1206,10 @@ TIFFFillTile(TIFF* tif, uint32 tile)
 	static const char module[] = "TIFFFillTile";
 	TIFFDirectory *td = &tif->tif_dir;

-        if (!_TIFFFillStriles( tif ) || !tif->tif_dir.td_stripbytecount)
-            return 0;
-
 	if ((tif->tif_flags&TIFF_NOREADRAW)==0)
 	{
-		uint64 bytecount = td->td_stripbytecount[tile];
-		if ((int64)bytecount <= 0) {
+		uint64 bytecount = TIFFGetStrileByteCount(tif, tile);
+		if( bytecount == 0 || bytecount > (uint64)TIFF_INT64_MAX ) {
 #if defined(__WIN32__) && (defined(_MSC_VER) || defined(__MINGW32__))
 			TIFFErrorExt(tif->tif_clientdata, module,
 				"%I64u: Invalid tile byte count, tile %lu",
@@ -1253,7 +1236,7 @@ TIFFFillTile(TIFF* tif, uint32 tile)
 			    (bytecount - 4096) / 10 > (uint64)stripsize  )
 			{
 				uint64 newbytecount = (uint64)stripsize * 10 + 4096;
-				if( (int64)newbytecount >= 0 )
+				if( newbytecount == 0 || newbytecount > (uint64)TIFF_INT64_MAX )
 				{
 #if defined(__WIN32__) && (defined(_MSC_VER) || defined(__MINGW32__))
 					TIFFWarningExt(tif->tif_clientdata, module,
@@ -1278,13 +1261,13 @@ TIFFFillTile(TIFF* tif, uint32 tile)
 			 * We must check for overflow, potentially causing
 			 * an OOB read. Instead of simple
 			 *
-			 *  td->td_stripoffset[tile]+bytecount > tif->tif_size
+			 *  TIFFGetStrileOffset(tif, tile)+bytecount > tif->tif_size
 			 *
 			 * comparison (which can overflow) we do the following
 			 * two comparisons:
 			 */
 			if (bytecount > (uint64)tif->tif_size ||
-			    td->td_stripoffset[tile] > (uint64)tif->tif_size - bytecount) {
+			    TIFFGetStrileOffset(tif, tile) > (uint64)tif->tif_size - bytecount) {
 				tif->tif_curtile = NOTILE;
 				return (0);
 			}
@@ -1313,7 +1296,7 @@ TIFFFillTile(TIFF* tif, uint32 tile)

 			tif->tif_rawdatasize = (tmsize_t)bytecount;
 			tif->tif_rawdata =
-				tif->tif_base + (tmsize_t)td->td_stripoffset[tile];
+				tif->tif_base + (tmsize_t)TIFFGetStrileOffset(tif, tile);
                        tif->tif_rawdataoff = 0;
                        tif->tif_rawdataloaded = (tmsize_t) bytecount;
 			tif->tif_flags |= TIFF_BUFFERMMAP;
@@ -1440,9 +1423,6 @@ TIFFStartStrip(TIFF* tif, uint32 strip)
 {
 	TIFFDirectory *td = &tif->tif_dir;

-        if (!_TIFFFillStriles( tif ) || !tif->tif_dir.td_stripbytecount)
-            return 0;
-
 	if ((tif->tif_flags & TIFF_CODERSETUP) == 0) {
 		if (!(*tif->tif_setupdecode)(tif))
 			return (0);
@@ -1463,10 +1443,18 @@ TIFFStartStrip(TIFF* tif, uint32 strip)
 		if( tif->tif_rawdataloaded > 0 )
 			tif->tif_rawcc = tif->tif_rawdataloaded;
 		else
-			tif->tif_rawcc = (tmsize_t)td->td_stripbytecount[strip];
+			tif->tif_rawcc = (tmsize_t)TIFFGetStrileByteCount(tif, strip);
 	}
-	return ((*tif->tif_predecode)(tif,
-			(uint16)(strip / td->td_stripsperimage)));
+	if ((*tif->tif_predecode)(tif,
+			(uint16)(strip / td->td_stripsperimage)) == 0 ) {
+            /* Needed for example for scanline access, if tif_predecode */
+            /* fails, and we try to read the same strip again. Without invalidating */
+            /* tif_curstrip, we'd call tif_decoderow() on a possibly invalid */
+            /* codec state. */
+            tif->tif_curstrip = NOSTRIP;
+            return 0;
+        }
+        return 1;
 }

 /*
@@ -1480,9 +1468,6 @@ TIFFStartTile(TIFF* tif, uint32 tile)
 	TIFFDirectory *td = &tif->tif_dir;
        uint32 howmany32;

-        if (!_TIFFFillStriles( tif ) || !tif->tif_dir.td_stripbytecount)
-                return 0;
-
 	if ((tif->tif_flags & TIFF_CODERSETUP) == 0) {
 		if (!(*tif->tif_setupdecode)(tif))
 			return (0);
@@ -1513,7 +1498,7 @@ TIFFStartTile(TIFF* tif, uint32 tile)
 		if( tif->tif_rawdataloaded > 0 )
 			tif->tif_rawcc = tif->tif_rawdataloaded;
 		else
-			tif->tif_rawcc = (tmsize_t)td->td_stripbytecount[tile];
+			tif->tif_rawcc = (tmsize_t)TIFFGetStrileByteCount(tif, tile);
 	}
 	return ((*tif->tif_predecode)(tif,
 			(uint16)(tile/td->td_stripsperimage)));
@@ -1528,13 +1513,100 @@ TIFFCheckRead(TIFF* tif, int tiles)
 	}
 	if (tiles ^ isTiled(tif)) {
 		TIFFErrorExt(tif->tif_clientdata, tif->tif_name, tiles ?
-		    "Can not read tiles from a stripped image" :
+		    "Can not read tiles from a striped image" :
 		    "Can not read scanlines from a tiled image");
 		return (0);
 	}
 	return (1);
 }

+/* Use the provided input buffer (inbuf, insize) and decompress it into
+ * (outbuf, outsize).
+ * This function replaces the use of TIFFReadEncodedStrip()/TIFFReadEncodedTile()
+ * when the user can provide the buffer for the input data, for example when
+ * he wants to avoid libtiff to read the strile offset/count values from the
+ * [Strip|Tile][Offsets/ByteCounts] array.
+ * inbuf content must be writable (if bit reversal is needed)
+ * Returns 1 in case of success, 0 otherwise.
+ */
+int      TIFFReadFromUserBuffer(TIFF* tif, uint32 strile,
+                                void* inbuf, tmsize_t insize,
+                                void* outbuf, tmsize_t outsize)
+{
+    static const char module[] = "TIFFReadFromUserBuffer";
+    TIFFDirectory *td = &tif->tif_dir;
+    int ret = 1;
+    uint32 old_tif_flags = tif->tif_flags;
+    tmsize_t old_rawdatasize = tif->tif_rawdatasize;
+    void* old_rawdata = tif->tif_rawdata;
+
+    if (tif->tif_mode == O_WRONLY) {
+        TIFFErrorExt(tif->tif_clientdata, tif->tif_name, "File not open for reading");
+        return 0;
+    }
+    if (tif->tif_flags&TIFF_NOREADRAW)
+    {
+        TIFFErrorExt(tif->tif_clientdata, module,
+                "Compression scheme does not support access to raw uncompressed data");
+        return 0;
+    }
+
+    tif->tif_flags &= ~TIFF_MYBUFFER;
+    tif->tif_flags |= TIFF_BUFFERMMAP;
+    tif->tif_rawdatasize = insize;
+    tif->tif_rawdata = inbuf;
+    tif->tif_rawdataoff = 0;
+    tif->tif_rawdataloaded = insize;
+
+    if (!isFillOrder(tif, td->td_fillorder) &&
+        (tif->tif_flags & TIFF_NOBITREV) == 0)
+    {
+        TIFFReverseBits(inbuf, insize);
+    }
+
+    if( TIFFIsTiled(tif) )
+    {
+        if( !TIFFStartTile(tif, strile) ||
+            !(*tif->tif_decodetile)(tif, (uint8*) outbuf, outsize, 
+                                    (uint16)(strile/td->td_stripsperimage)) )
+        {
+            ret = 0;
+        }
+    }
+    else
+    {
+        uint32 rowsperstrip=td->td_rowsperstrip;
+        uint32 stripsperplane;
+        if (rowsperstrip>td->td_imagelength)
+            rowsperstrip=td->td_imagelength;
+        stripsperplane= TIFFhowmany_32_maxuint_compat(td->td_imagelength, rowsperstrip);
+        if( !TIFFStartStrip(tif, strile) ||
+            !(*tif->tif_decodestrip)(tif, (uint8*) outbuf, outsize, 
+                                     (uint16)(strile/stripsperplane)) )
+        {
+            ret = 0;
+        }
+    }
+    if( ret )
+    {
+        (*tif->tif_postdecode)(tif, (uint8*) outbuf, outsize);
+    }
+
+    if (!isFillOrder(tif, td->td_fillorder) &&
+        (tif->tif_flags & TIFF_NOBITREV) == 0)
+    {
+        TIFFReverseBits(inbuf, insize);
+    }
+
+    tif->tif_flags = old_tif_flags;
+    tif->tif_rawdatasize = old_rawdatasize;
+    tif->tif_rawdata = old_rawdata;
+    tif->tif_rawdataoff = 0;
+    tif->tif_rawdataloaded = 0;
+
+    return ret;
+}
+
 void
 _TIFFNoPostDecode(TIFF* tif, uint8* buf, tmsize_t cc)
 {
@@ -129,15 +129,8 @@ TIFFVStripSize(TIFF* tif, uint32 nrows)
 {
 	static const char module[] = "TIFFVStripSize";
 	uint64 m;
-	tmsize_t n;
 	m=TIFFVStripSize64(tif,nrows);
-	n=(tmsize_t)m;
-	if ((uint64)n!=m)
-	{
-		TIFFErrorExt(tif->tif_clientdata,module,"Integer overflow");
-		n=0;
-	}
-	return(n);
+        return _TIFFCastUInt64ToSSize(tif, m, module);
 }

 /*
@@ -147,8 +140,7 @@ uint64
 TIFFRawStripSize64(TIFF* tif, uint32 strip)
 {
 	static const char module[] = "TIFFRawStripSize64";
-	TIFFDirectory* td = &tif->tif_dir;
-	uint64 bytecount = td->td_stripbytecount[strip];
+	uint64 bytecount = TIFFGetStrileByteCount(tif, strip);

 	if (bytecount == 0)
 	{
@@ -211,15 +203,8 @@ TIFFStripSize(TIFF* tif)
 {
 	static const char module[] = "TIFFStripSize";
 	uint64 m;
-	tmsize_t n;
 	m=TIFFStripSize64(tif);
-	n=(tmsize_t)m;
-	if ((uint64)n!=m)
-	{
-		TIFFErrorExt(tif->tif_clientdata,module,"Integer overflow");
-		n=0;
-	}
-	return(n);
+	return _TIFFCastUInt64ToSSize(tif, m, module);
 }

 /*
@@ -330,14 +315,8 @@ TIFFScanlineSize(TIFF* tif)
 {
 	static const char module[] = "TIFFScanlineSize";
 	uint64 m;
-	tmsize_t n;
 	m=TIFFScanlineSize64(tif);
-	n=(tmsize_t)m;
-	if ((uint64)n!=m) {
-		TIFFErrorExt(tif->tif_clientdata,module,"Integer arithmetic overflow");
-		n=0;
-	}
-	return(n);
+	return _TIFFCastUInt64ToSSize(tif, m, module);
 }

 /*
@@ -366,15 +345,8 @@ TIFFRasterScanlineSize(TIFF* tif)
 {
 	static const char module[] = "TIFFRasterScanlineSize";
 	uint64 m;
-	tmsize_t n;
 	m=TIFFRasterScanlineSize64(tif);
-	n=(tmsize_t)m;
-	if ((uint64)n!=m)
-	{
-		TIFFErrorExt(tif->tif_clientdata,module,"Integer arithmetic overflow");
-		n=0;
-	}
-	return(n);
+	return _TIFFCastUInt64ToSSize(tif, m, module);
 }

 /* vim: set ts=8 sts=8 sw=8 noet: */
@@ -122,17 +122,17 @@ ThunderDecode(TIFF* tif, uint8* op, tmsize_t maxpixels)
 			break;
 		case THUNDER_2BITDELTAS:	/* 2-bit deltas */
 			if ((delta = ((n >> 4) & 3)) != DELTA2_SKIP)
-				SETPIXEL(op, lastpixel + twobitdeltas[delta]);
+				SETPIXEL(op, (unsigned)((int)lastpixel + twobitdeltas[delta]));
 			if ((delta = ((n >> 2) & 3)) != DELTA2_SKIP)
-				SETPIXEL(op, lastpixel + twobitdeltas[delta]);
+				SETPIXEL(op, (unsigned)((int)lastpixel + twobitdeltas[delta]));
 			if ((delta = (n & 3)) != DELTA2_SKIP)
-				SETPIXEL(op, lastpixel + twobitdeltas[delta]);
+				SETPIXEL(op, (unsigned)((int)lastpixel + twobitdeltas[delta]));
 			break;
 		case THUNDER_3BITDELTAS:	/* 3-bit deltas */
 			if ((delta = ((n >> 3) & 7)) != DELTA3_SKIP)
-				SETPIXEL(op, lastpixel + threebitdeltas[delta]);
+				SETPIXEL(op, (unsigned)((int)lastpixel + threebitdeltas[delta]));
 			if ((delta = (n & 7)) != DELTA3_SKIP)
-				SETPIXEL(op, lastpixel + threebitdeltas[delta]);
+				SETPIXEL(op, (unsigned)((int)lastpixel + threebitdeltas[delta]));
 			break;
 		case THUNDER_RAW:		/* raw data */
 			SETPIXEL(op, n);
@@ -181,15 +181,8 @@ TIFFTileRowSize(TIFF* tif)
 {
 	static const char module[] = "TIFFTileRowSize";
 	uint64 m;
-	tmsize_t n;
 	m=TIFFTileRowSize64(tif);
-	n=(tmsize_t)m;
-	if ((uint64)n!=m)
-	{
-		TIFFErrorExt(tif->tif_clientdata,module,"Integer overflow");
-		n=0;
-	}
-	return(n);
+	return _TIFFCastUInt64ToSSize(tif, m, module);
 }

 /*
@@ -248,15 +241,8 @@ TIFFVTileSize(TIFF* tif, uint32 nrows)
 {
 	static const char module[] = "TIFFVTileSize";
 	uint64 m;
-	tmsize_t n;
 	m=TIFFVTileSize64(tif,nrows);
-	n=(tmsize_t)m;
-	if ((uint64)n!=m)
-	{
-		TIFFErrorExt(tif->tif_clientdata,module,"Integer overflow");
-		n=0;
-	}
-	return(n);
+	return _TIFFCastUInt64ToSSize(tif, m, module);
 }

 /*
@@ -272,15 +258,8 @@ TIFFTileSize(TIFF* tif)
 {
 	static const char module[] = "TIFFTileSize";
 	uint64 m;
-	tmsize_t n;
 	m=TIFFTileSize64(tif);
-	n=(tmsize_t)m;
-	if ((uint64)n!=m)
-	{
-		TIFFErrorExt(tif->tif_clientdata,module,"Integer overflow");
-		n=0;
-	}
-	return(n);
+	return _TIFFCastUInt64ToSSize(tif, m, module);
 }

 /*
@@ -162,7 +162,7 @@ _tiffMapProc(thandle_t fd, void** pbase, toff_t* psize)
 {
 	uint64 size64 = _tiffSizeProc(fd);
 	tmsize_t sizem = (tmsize_t)size64;
-	if ((uint64)sizem==size64) {
+	if (size64 && (uint64)sizem==size64) {
 		fd_as_handle_union_t fdh;
 		fdh.h = fd;
 		*pbase = (void*)
@@ -267,6 +267,12 @@ TWebPPreDecode(TIFF* tif, uint16 s)
      segment_height = td->td_rowsperstrip;
  }

+  if( segment_width > 16383 || segment_height > 16383 ) {
+      TIFFErrorExt(tif->tif_clientdata, module,
+                   "WEBP maximum image dimensions are 16383 x 16383.");
+      return 0;
+  }
+
  if( (sp->state & LSTATE_INIT_DECODE) == 0 )
      tif->tif_setupdecode(tif);
      
@@ -333,7 +339,7 @@ TWebPSetupEncode(TIFF* tif)
  }
  
  /* check bits per sample and data type */
-  if ((nBitsPerSample != 8) && (sampleFormat != 1)) {
+  if ((nBitsPerSample != 8) || (sampleFormat != SAMPLEFORMAT_UINT)) {
    TIFFErrorExt(tif->tif_clientdata, module,
                "WEBP driver requires 8 bit unsigned data");
    return 0;
@@ -356,7 +362,7 @@ TWebPSetupEncode(TIFF* tif)
  }

  if (!WebPConfigInitInternal(&sp->sEncoderConfig, WEBP_PRESET_DEFAULT,
-                              sp->quality_level,
+                              (float)sp->quality_level,
                              WEBP_ENCODER_ABI_VERSION)) {
    TIFFErrorExt(tif->tif_clientdata, module,
      "Error creating WebP encoder configuration.");
@@ -579,7 +585,7 @@ TWebPVSetField(TIFF* tif, uint32 tag, va_list ap)
    #if WEBP_ENCODER_ABI_VERSION >= 0x0100
    sp->lossless = va_arg(ap, int);
    if (sp->lossless){
-      sp->quality_level = 100.0f;      
+      sp->quality_level = 100;
    }
    return 1;
    #else
@@ -628,6 +634,7 @@ TIFFInitWebP(TIFF* tif, int scheme)
  static const char module[] = "TIFFInitWebP";
  WebPState* sp;

+  (void)scheme;
  assert( scheme == COMPRESSION_WEBP );

  /*
@@ -656,7 +663,7 @@ TIFFInitWebP(TIFF* tif, int scheme)
  tif->tif_tagmethods.vsetfield = TWebPVSetField;	/* hook for codec tags */

  /* Default values for codec-specific fields */
-  sp->quality_level = 75.0f;		/* default comp. level */
+  sp->quality_level = 75;		/* default comp. level */
  sp->lossless = 0; /* default to false */
  sp->state = 0;
  sp->nSamples = 0;
@@ -27,34 +27,38 @@
 * Scott Wagner (wagner@itek.com), Itek Graphix, Rochester, NY USA
 */

-/*
-  CreateFileA/CreateFileW return type 'HANDLE'.
-
-  thandle_t is declared like
-
-    DECLARE_HANDLE(thandle_t);
-
-  in tiffio.h.
-
-  Windows (from winnt.h) DECLARE_HANDLE logic looks like
-
-  #ifdef STRICT
-    typedef void *HANDLE;
-  #define DECLARE_HANDLE(name) struct name##__ { int unused; }; typedef struct name##__ *name
-  #else
-    typedef PVOID HANDLE;
-  #define DECLARE_HANDLE(name) typedef HANDLE name
-  #endif
-
-  See http://bugzilla.maptools.org/show_bug.cgi?id=1941 for problems in WIN64
-  builds resulting from this.  Unfortunately, the proposed patch was lost.
-
-*/
-  
 #include "tiffiop.h"

 #include <windows.h>

+/*
+  CreateFileA/CreateFileW return type 'HANDLE' while TIFFFdOpen() takes 'int',
+  which is formally incompatible and can even seemingly be of different size:
+  HANDLE is 64 bit under Win64, while int is still 32 bits there.
+
+  However, only the lower 32 bits of a HANDLE are significant under Win64 as,
+  for interoperability reasons, they must have the same values in 32- and
+  64-bit programs running on the same system, see
+
+  https://docs.microsoft.com/en-us/windows/win32/winprog64/interprocess-communication
+
+  Because of this, it is safe to define the following trivial functions for
+  casting between ints and HANDLEs, which are only really needed to avoid
+  compiler warnings (and, perhaps, to make the code slightly more clear).
+  Note that using the intermediate cast to "intptr_t" is crucial for warning
+  avoidance, as this integer type has the same size as HANDLE in all builds.
+*/
+
+static inline thandle_t thandle_from_int(int ifd)
+{
+    return (thandle_t)(intptr_t)ifd;
+}
+
+static inline int thandle_to_int(thandle_t fd)
+{
+    return (int)(intptr_t)fd;
+}
+
 static tmsize_t
 _tiffReadProc(thandle_t fd, void* buf, tmsize_t size)
 {
@@ -151,9 +155,11 @@ _tiffCloseProc(thandle_t fd)
 static uint64
 _tiffSizeProc(thandle_t fd)
 {
-	ULARGE_INTEGER m;
-	m.LowPart=GetFileSize(fd,&m.HighPart);
-	return(m.QuadPart);
+	LARGE_INTEGER m;
+	if (GetFileSizeEx(fd,&m))
+		return(m.QuadPart);
+	else
+		return(0);
 }

 static int
@@ -185,7 +191,7 @@ _tiffMapProc(thandle_t fd, void** pbase, toff_t* psize)

 	size = _tiffSizeProc(fd);
 	sizem = (tmsize_t)size;
-	if ((uint64)sizem!=size)
+	if (!size || (uint64)sizem!=size)
 		return (0);

 	/* By passing in 0 for the maximum file size, it specifies that we
@@ -237,7 +243,7 @@ TIFFFdOpen(int ifd, const char* name, const char* mode)
 			break;
 		}
 	}
-	tif = TIFFClientOpen(name, mode, (thandle_t)ifd, /* FIXME: WIN64 cast to pointer warning */
+	tif = TIFFClientOpen(name, mode, thandle_from_int(ifd),
 			_tiffReadProc, _tiffWriteProc,
 			_tiffSeekProc, _tiffCloseProc, _tiffSizeProc,
 			fSuppressMap ? _tiffDummyMapProc : _tiffMapProc,
@@ -282,7 +288,7 @@ TIFFOpen(const char* name, const char* mode)
 		return ((TIFF *)0);
 	}

-	tif = TIFFFdOpen((int)fd, name, mode);   /* FIXME: WIN64 cast from pointer to int warning */
+	tif = TIFFFdOpen(thandle_to_int(fd), name, mode);
 	if(!tif)
 		CloseHandle(fd);
 	return tif;
@@ -337,7 +343,7 @@ TIFFOpenW(const wchar_t* name, const char* mode)
 				    NULL, NULL);
 	}

-	tif = TIFFFdOpen((int)fd,    /* FIXME: WIN64 cast from pointer to int warning */
+	tif = TIFFFdOpen(thandle_to_int(fd),
 			 (mbname != NULL) ? mbname : "<unknown>", mode);
 	if(!tif)
 		CloseHandle(fd);
@@ -128,10 +128,10 @@ TIFFWriteScanline(TIFF* tif, void* buf, uint32 row, uint16 sample)
 		tif->tif_rawcc = 0;
 		tif->tif_rawcp = tif->tif_rawdata;

-		if( td->td_stripbytecount[strip] > 0 )
+		if( td->td_stripbytecount_p[strip] > 0 )
 		{
 			/* if we are writing over existing tiles, zero length */
-			td->td_stripbytecount[strip] = 0;
+			td->td_stripbytecount_p[strip] = 0;

 			/* this forces TIFFAppendToStrip() to do a seek */
 			tif->tif_curoff = 0;
@@ -176,6 +176,32 @@ TIFFWriteScanline(TIFF* tif, void* buf, uint32 row, uint16 sample)
 	return (status);
 }

+/* Make sure that at the first attempt of rewriting a tile/strip, we will have */
+/* more bytes available in the output buffer than the previous byte count, */
+/* so that TIFFAppendToStrip() will detect the overflow when it is called the first */
+/* time if the new compressed tile is bigger than the older one. (GDAL #4771) */
+static int _TIFFReserveLargeEnoughWriteBuffer(TIFF* tif, uint32 strip_or_tile)
+{
+    TIFFDirectory *td = &tif->tif_dir;
+    if( td->td_stripbytecount_p[strip_or_tile] > 0 )
+    {
+        /* The +1 is to ensure at least one extra bytes */
+        /* The +4 is because the LZW encoder flushes 4 bytes before the limit */
+        uint64 safe_buffer_size = (uint64)(td->td_stripbytecount_p[strip_or_tile] + 1 + 4);
+        if( tif->tif_rawdatasize <= (tmsize_t)safe_buffer_size )
+        {
+            if( !(TIFFWriteBufferSetup(tif, NULL,
+                (tmsize_t)TIFFroundup_64(safe_buffer_size, 1024))) )
+                return 0;
+        }
+
+        /* Force TIFFAppendToStrip() to consider placing data at end
+            of file. */
+        tif->tif_curoff = 0;
+    }
+    return 1;
+}
+
 /*
 * Encode the supplied data and write it to the
 * specified strip.
@@ -222,6 +248,13 @@ TIFFWriteEncodedStrip(TIFF* tif, uint32 strip, void* data, tmsize_t cc)
        tif->tif_flags |= TIFF_BUF4WRITE;
 	tif->tif_curstrip = strip;

+	if( !_TIFFReserveLargeEnoughWriteBuffer(tif, strip) ) {
+            return ((tmsize_t)(-1));
+        }
+
+        tif->tif_rawcc = 0;
+        tif->tif_rawcp = tif->tif_rawdata;
+
        if (td->td_stripsperimage == 0) {
                TIFFErrorExt(tif->tif_clientdata, module, "Zero strips per image");
                return ((tmsize_t) -1);
@@ -234,27 +267,6 @@ TIFFWriteEncodedStrip(TIFF* tif, uint32 strip, void* data, tmsize_t cc)
 		tif->tif_flags |= TIFF_CODERSETUP;
 	}

-	if( td->td_stripbytecount[strip] > 0 )
-        {
-            /* Make sure that at the first attempt of rewriting the tile, we will have */
-            /* more bytes available in the output buffer than the previous byte count, */
-            /* so that TIFFAppendToStrip() will detect the overflow when it is called the first */
-            /* time if the new compressed tile is bigger than the older one. (GDAL #4771) */
-            if( tif->tif_rawdatasize <= (tmsize_t)td->td_stripbytecount[strip] )
-            {
-                if( !(TIFFWriteBufferSetup(tif, NULL,
-                    (tmsize_t)TIFFroundup_64((uint64)(td->td_stripbytecount[strip] + 1), 1024))) )
-                    return ((tmsize_t)(-1));
-            }
-
-	    /* Force TIFFAppendToStrip() to consider placing data at end
-               of file. */
-            tif->tif_curoff = 0;
-        }
-
-    tif->tif_rawcc = 0;
-    tif->tif_rawcp = tif->tif_rawdata;
-
 	tif->tif_flags &= ~TIFF_POSTENCODE;

    /* shortcut to avoid an extra memcpy() */
@@ -402,22 +414,8 @@ TIFFWriteEncodedTile(TIFF* tif, uint32 tile, void* data, tmsize_t cc)
        tif->tif_flags |= TIFF_BUF4WRITE;
 	tif->tif_curtile = tile;

-	if( td->td_stripbytecount[tile] > 0 )
-        {
-            /* Make sure that at the first attempt of rewriting the tile, we will have */
-            /* more bytes available in the output buffer than the previous byte count, */
-            /* so that TIFFAppendToStrip() will detect the overflow when it is called the first */
-            /* time if the new compressed tile is bigger than the older one. (GDAL #4771) */
-            if( tif->tif_rawdatasize <= (tmsize_t) td->td_stripbytecount[tile] )
-            {
-                if( !(TIFFWriteBufferSetup(tif, NULL,
-                    (tmsize_t)TIFFroundup_64((uint64)(td->td_stripbytecount[tile] + 1), 1024))) )
-                    return ((tmsize_t)(-1));
-            }
-
-	    /* Force TIFFAppendToStrip() to consider placing data at end
-               of file. */
-            tif->tif_curoff = 0;
+        if( !_TIFFReserveLargeEnoughWriteBuffer(tif, tile) ) {
+            return ((tmsize_t)(-1));
        }

 	tif->tif_rawcc = 0;
@@ -535,22 +533,29 @@ TIFFSetupStrips(TIFF* tif)
 		    isUnspecified(tif, FIELD_ROWSPERSTRIP) ?
 			td->td_samplesperpixel : TIFFNumberOfStrips(tif);
 	td->td_nstrips = td->td_stripsperimage;
+        /* TIFFWriteDirectoryTagData has a limitation to 0x80000000U bytes */
+        if( td->td_nstrips >= 0x80000000U / ((tif->tif_flags&TIFF_BIGTIFF)?0x8U:0x4U) )
+        {
+            TIFFErrorExt(tif->tif_clientdata, "TIFFSetupStrips",
+                         "Too large Strip/Tile Offsets/ByteCounts arrays");
+            return 0;
+        }
 	if (td->td_planarconfig == PLANARCONFIG_SEPARATE)
 		td->td_stripsperimage /= td->td_samplesperpixel;
-	td->td_stripoffset = (uint64 *)
+	td->td_stripoffset_p = (uint64 *)
            _TIFFCheckMalloc(tif, td->td_nstrips, sizeof (uint64),
                             "for \"StripOffsets\" array");
-	td->td_stripbytecount = (uint64 *)
+	td->td_stripbytecount_p = (uint64 *)
            _TIFFCheckMalloc(tif, td->td_nstrips, sizeof (uint64),
                             "for \"StripByteCounts\" array");
-	if (td->td_stripoffset == NULL || td->td_stripbytecount == NULL)
+	if (td->td_stripoffset_p == NULL || td->td_stripbytecount_p == NULL)
 		return (0);
 	/*
 	 * Place data at the end-of-file
 	 * (by setting offsets to zero).
 	 */
-	_TIFFmemset(td->td_stripoffset, 0, td->td_nstrips*sizeof (uint64));
-	_TIFFmemset(td->td_stripbytecount, 0, td->td_nstrips*sizeof (uint64));
+	_TIFFmemset(td->td_stripoffset_p, 0, td->td_nstrips*sizeof (uint64));
+	_TIFFmemset(td->td_stripbytecount_p, 0, td->td_nstrips*sizeof (uint64));
 	TIFFSetFieldBit(tif, FIELD_STRIPOFFSETS);
 	TIFFSetFieldBit(tif, FIELD_STRIPBYTECOUNTS);
 	return (1);
@@ -572,7 +577,7 @@ TIFFWriteCheck(TIFF* tif, int tiles, const char* module)
 	}
 	if (tiles ^ isTiled(tif)) {
 		TIFFErrorExt(tif->tif_clientdata, module, tiles ?
-		    "Can not write tiles to a stripped image" :
+		    "Can not write tiles to a striped image" :
 		    "Can not write scanlines to a tiled image");
 		return (0);
 	}
@@ -610,7 +615,7 @@ TIFFWriteCheck(TIFF* tif, int tiles, const char* module)
 			return (0);
 		}
 	}
-	if (tif->tif_dir.td_stripoffset == NULL && !TIFFSetupStrips(tif)) {
+	if (tif->tif_dir.td_stripoffset_p == NULL && !TIFFSetupStrips(tif)) {
 		tif->tif_dir.td_nstrips = 0;
 		TIFFErrorExt(tif->tif_clientdata, module, "No space for %s arrays",
 		    isTiled(tif) ? "tile" : "strip");
@@ -628,6 +633,20 @@ TIFFWriteCheck(TIFF* tif, int tiles, const char* module)
 	if (tif->tif_scanlinesize == 0)
 		return (0);
 	tif->tif_flags |= TIFF_BEENWRITING;
+
+        if( tif->tif_dir.td_stripoffset_entry.tdir_tag != 0 &&
+            tif->tif_dir.td_stripoffset_entry.tdir_count == 0 &&
+            tif->tif_dir.td_stripoffset_entry.tdir_type == 0 &&
+            tif->tif_dir.td_stripoffset_entry.tdir_offset.toff_long8 == 0 &&
+            tif->tif_dir.td_stripbytecount_entry.tdir_tag != 0 &&
+            tif->tif_dir.td_stripbytecount_entry.tdir_count == 0 &&
+            tif->tif_dir.td_stripbytecount_entry.tdir_type == 0 &&
+            tif->tif_dir.td_stripbytecount_entry.tdir_offset.toff_long8 == 0 &&
+            !(tif->tif_flags & TIFF_DIRTYDIRECT)  )
+        {
+            TIFFForceStrileArrayWriting(tif);
+        }
+
 	return (1);
 }

@@ -649,6 +668,10 @@ TIFFWriteBufferSetup(TIFF* tif, void* bp, tmsize_t size)
 	if (size == (tmsize_t)(-1)) {
 		size = (isTiled(tif) ?
 		    tif->tif_tilesize : TIFFStripSize(tif));
+
+                /* Adds 10% margin for cases where compression would expand a bit */
+                if( size < TIFF_TMSIZE_T_MAX - size / 10 )
+                    size += size / 10;
 		/*
 		 * Make raw data buffer at least 8K
 		 */
@@ -684,9 +707,9 @@ TIFFGrowStrips(TIFF* tif, uint32 delta, const char* module)
 	uint64* new_stripbytecount;

 	assert(td->td_planarconfig == PLANARCONFIG_CONTIG);
-	new_stripoffset = (uint64*)_TIFFrealloc(td->td_stripoffset,
+	new_stripoffset = (uint64*)_TIFFrealloc(td->td_stripoffset_p,
 		(td->td_nstrips + delta) * sizeof (uint64));
-	new_stripbytecount = (uint64*)_TIFFrealloc(td->td_stripbytecount,
+	new_stripbytecount = (uint64*)_TIFFrealloc(td->td_stripbytecount_p,
 		(td->td_nstrips + delta) * sizeof (uint64));
 	if (new_stripoffset == NULL || new_stripbytecount == NULL) {
 		if (new_stripoffset)
@@ -697,11 +720,11 @@ TIFFGrowStrips(TIFF* tif, uint32 delta, const char* module)
 		TIFFErrorExt(tif->tif_clientdata, module, "No space to expand strip arrays");
 		return (0);
 	}
-	td->td_stripoffset = new_stripoffset;
-	td->td_stripbytecount = new_stripbytecount;
-	_TIFFmemset(td->td_stripoffset + td->td_nstrips,
+	td->td_stripoffset_p = new_stripoffset;
+	td->td_stripbytecount_p = new_stripbytecount;
+	_TIFFmemset(td->td_stripoffset_p + td->td_nstrips,
 		    0, delta*sizeof (uint64));
-	_TIFFmemset(td->td_stripbytecount + td->td_nstrips,
+	_TIFFmemset(td->td_stripbytecount_p + td->td_nstrips,
 		    0, delta*sizeof (uint64));
 	td->td_nstrips += delta;
        tif->tif_flags |= TIFF_DIRTYDIRECT;
@@ -720,12 +743,12 @@ TIFFAppendToStrip(TIFF* tif, uint32 strip, uint8* data, tmsize_t cc)
 	uint64 m;
        int64 old_byte_count = -1;

-	if (td->td_stripoffset[strip] == 0 || tif->tif_curoff == 0) {
+	if (td->td_stripoffset_p[strip] == 0 || tif->tif_curoff == 0) {
            assert(td->td_nstrips > 0);

-            if( td->td_stripbytecount[strip] != 0 
-                && td->td_stripoffset[strip] != 0 
-                && td->td_stripbytecount[strip] >= (uint64) cc )
+            if( td->td_stripbytecount_p[strip] != 0 
+                && td->td_stripoffset_p[strip] != 0 
+                && td->td_stripbytecount_p[strip] >= (uint64) cc )
            {
                /* 
                 * There is already tile data on disk, and the new tile
@@ -734,7 +757,7 @@ TIFFAppendToStrip(TIFF* tif, uint32 strip, uint8* data, tmsize_t cc)
                 * more data to append to this strip before we are done
                 * depending on how we are getting called.
                 */
-                if (!SeekOK(tif, td->td_stripoffset[strip])) {
+                if (!SeekOK(tif, td->td_stripoffset_p[strip])) {
                    TIFFErrorExt(tif->tif_clientdata, module,
                                 "Seek error at scanline %lu",
                                 (unsigned long)tif->tif_row);
@@ -747,17 +770,17 @@ TIFFAppendToStrip(TIFF* tif, uint32 strip, uint8* data, tmsize_t cc)
                 * Seek to end of file, and set that as our location to 
                 * write this strip.
                 */
-                td->td_stripoffset[strip] = TIFFSeekFile(tif, 0, SEEK_END);
+                td->td_stripoffset_p[strip] = TIFFSeekFile(tif, 0, SEEK_END);
                tif->tif_flags |= TIFF_DIRTYSTRIP;
            }

-            tif->tif_curoff = td->td_stripoffset[strip];
+            tif->tif_curoff = td->td_stripoffset_p[strip];

            /*
             * We are starting a fresh strip/tile, so set the size to zero.
             */
-            old_byte_count = td->td_stripbytecount[strip];
-            td->td_stripbytecount[strip] = 0;
+            old_byte_count = td->td_stripbytecount_p[strip];
+            td->td_stripbytecount_p[strip] = 0;
 	}

 	m = tif->tif_curoff+cc;
@@ -774,9 +797,9 @@ TIFFAppendToStrip(TIFF* tif, uint32 strip, uint8* data, tmsize_t cc)
 		    return (0);
 	}
 	tif->tif_curoff = m;
-	td->td_stripbytecount[strip] += cc;
+	td->td_stripbytecount_p[strip] += cc;

-        if( (int64) td->td_stripbytecount[strip] != old_byte_count )
+        if( (int64) td->td_stripbytecount_p[strip] != old_byte_count )
            tif->tif_flags |= TIFF_DIRTYSTRIP;
            
 	return (1);
@@ -29,24 +29,22 @@
 *
 * ZIP (aka Deflate) Compression Support
 *
- * This file is simply an interface to the zlib library written by
+ * This file is an interface to the zlib library written by
 * Jean-loup Gailly and Mark Adler.  You must use version 1.0 or later
- * of the library: this code assumes the 1.0 API and also depends on
- * the ability to write the zlib header multiple times (one per strip)
- * which was not possible with versions prior to 0.95.  Note also that
- * older versions of this codec avoided this bug by suppressing the header
- * entirely.  This means that files written with the old library cannot
- * be read; they should be converted to a different compression scheme
- * and then reconverted.
+ * of the library.
 *
- * The data format used by the zlib library is described in the files
- * zlib-3.1.doc, deflate-1.1.doc and gzip-4.1.doc, available in the
- * directory ftp://ftp.uu.net/pub/archiving/zip/doc.  The library was
- * last found at ftp://ftp.uu.net/pub/archiving/zip/zlib/zlib-0.99.tar.gz.
+ * Optionally, libdeflate (https://github.com/ebiggers/libdeflate) may be used
+ * to do the compression and decompression, but only for whole strips and tiles.
+ * For scanline access, zlib will be sued as a fallback.
 */
 #include "tif_predict.h"
 #include "zlib.h"

+#if LIBDEFLATE_SUPPORT
+#include "libdeflate.h"
+#endif
+#define LIBDEFLATE_MAX_COMPRESSION_LEVEL 12
+
 #include <stdio.h>

 /*
@@ -70,6 +68,12 @@ typedef struct {
        z_stream        stream;
 	int             zipquality;            /* compression level */
 	int             state;                 /* state flags */
+	int             subcodec;              /* DEFLATE_SUBCODEC_ZLIB or DEFLATE_SUBCODEC_LIBDEFLATE */
+#if LIBDEFLATE_SUPPORT
+	int             libdeflate_state;       /* -1 = until first time ZIPEncode() / ZIPDecode() is called, 0 = use zlib, 1 = use libdeflate */
+	struct libdeflate_decompressor* libdeflate_dec;
+	struct libdeflate_compressor*   libdeflate_enc;
+#endif
 #define ZSTATE_INIT_DECODE 0x01
 #define ZSTATE_INIT_ENCODE 0x02

@@ -132,6 +136,9 @@ ZIPPreDecode(TIFF* tif, uint16 s)
 	if( (sp->state & ZSTATE_INIT_DECODE) == 0 )
            tif->tif_setupdecode( tif );

+#if LIBDEFLATE_SUPPORT
+        sp->libdeflate_state = -1;
+#endif
 	sp->stream.next_in = tif->tif_rawdata;
 	assert(sizeof(sp->stream.avail_in)==4);  /* if this assert gets raised,
 	    we need to simplify this code to reflect a ZLib that is likely updated
@@ -151,6 +158,77 @@ ZIPDecode(TIFF* tif, uint8* op, tmsize_t occ, uint16 s)
 	assert(sp != NULL);
 	assert(sp->state == ZSTATE_INIT_DECODE);

+#if LIBDEFLATE_SUPPORT
+        if( sp->libdeflate_state == 1 )
+            return 0;
+
+        /* If we have libdeflate support and we are asked to read a whole */
+        /* strip/tile, then go for using it */
+        do {
+            TIFFDirectory *td = &tif->tif_dir;
+
+            if( sp->libdeflate_state == 0 )
+                break;
+            if( sp->subcodec == DEFLATE_SUBCODEC_ZLIB )
+                break;
+
+            /* Check if we are in the situation where we can use libdeflate */
+            if (isTiled(tif)) {
+                if( TIFFTileSize64(tif) != (uint64)occ )
+                    break;
+            } else {
+                uint32 strip_height = td->td_imagelength - tif->tif_row;
+                if (strip_height > td->td_rowsperstrip)
+                    strip_height = td->td_rowsperstrip;
+                if( TIFFVStripSize64(tif, strip_height) != (uint64)occ )
+                    break;
+            }
+
+            /* Check for overflow */
+            if( (size_t)tif->tif_rawcc != (uint64)tif->tif_rawcc )
+                break;
+            if( (size_t)occ != (uint64)occ )
+                break;
+
+            /* Go for decompression using libdeflate */
+            {
+                enum libdeflate_result res;
+                if( sp->libdeflate_dec == NULL )
+                {
+                    sp->libdeflate_dec = libdeflate_alloc_decompressor();
+                    if( sp->libdeflate_dec == NULL )
+                    {
+                        break;
+                    }
+                }
+
+                sp->libdeflate_state = 1;
+
+                res = libdeflate_zlib_decompress(
+                    sp->libdeflate_dec, tif->tif_rawcp, (size_t)tif->tif_rawcc, op, (size_t)occ, NULL);
+
+                tif->tif_rawcp += tif->tif_rawcc;
+                tif->tif_rawcc = 0;
+
+                /* We accept LIBDEFLATE_INSUFFICIENT_SPACE has a return */
+                /* There are odd files in the wild where the last strip, when */
+                /* it is smaller in height than td_rowsperstrip, actually contains */
+                /* data for td_rowsperstrip lines. Just ignore that silently. */
+                if( res != LIBDEFLATE_SUCCESS &&
+                    res != LIBDEFLATE_INSUFFICIENT_SPACE )
+                {
+                    TIFFErrorExt(tif->tif_clientdata, module,
+                                 "Decoding error at scanline %lu",
+                                 (unsigned long) tif->tif_row);
+                    return 0;
+                }
+
+                return 1;
+            }
+        } while(0);
+        sp->libdeflate_state = 0;
+#endif /* LIBDEFLATE_SUPPORT */
+
        sp->stream.next_in = tif->tif_rawcp;
        
 	sp->stream.next_out = op;
@@ -198,6 +276,7 @@ ZIPSetupEncode(TIFF* tif)
 {
 	static const char module[] = "ZIPSetupEncode";
 	ZIPState* sp = EncoderState(tif);
+        int cappedQuality;

 	assert(sp != NULL);
 	if (sp->state & ZSTATE_INIT_DECODE) {
@@ -205,7 +284,11 @@ ZIPSetupEncode(TIFF* tif)
 		sp->state = 0;
 	}

-	if (deflateInit(&sp->stream, sp->zipquality) != Z_OK) {
+        cappedQuality = sp->zipquality;
+        if( cappedQuality > Z_BEST_COMPRESSION )
+            cappedQuality = Z_BEST_COMPRESSION;
+
+	if (deflateInit(&sp->stream, cappedQuality) != Z_OK) {
 		TIFFErrorExt(tif->tif_clientdata, module, "%s", SAFE_MSG(sp));
 		return (0);
 	} else {
@@ -227,6 +310,9 @@ ZIPPreEncode(TIFF* tif, uint16 s)
 	if( sp->state != ZSTATE_INIT_ENCODE )
            tif->tif_setupencode( tif );

+#if LIBDEFLATE_SUPPORT
+        sp->libdeflate_state = -1;
+#endif
 	sp->stream.next_out = tif->tif_rawdata;
 	assert(sizeof(sp->stream.avail_out)==4);  /* if this assert gets raised,
 	    we need to simplify this code to reflect a ZLib that is likely updated
@@ -249,6 +335,95 @@ ZIPEncode(TIFF* tif, uint8* bp, tmsize_t cc, uint16 s)
 	assert(sp->state == ZSTATE_INIT_ENCODE);

 	(void) s;
+
+#if LIBDEFLATE_SUPPORT
+        if( sp->libdeflate_state == 1 )
+            return 0;
+
+        /* If we have libdeflate support and we are asked to write a whole */
+        /* strip/tile, then go for using it */
+        do {
+            TIFFDirectory *td = &tif->tif_dir;
+
+            if( sp->libdeflate_state == 0 )
+                break;
+            if( sp->subcodec == DEFLATE_SUBCODEC_ZLIB )
+                break;
+
+            /* Libdeflate does not support the 0-compression level */
+            if( sp->zipquality == Z_NO_COMPRESSION )
+                break;
+
+            /* Check if we are in the situation where we can use libdeflate */
+            if (isTiled(tif)) {
+                if( TIFFTileSize64(tif) != (uint64)cc )
+                    break;
+            } else {
+                uint32 strip_height = td->td_imagelength - tif->tif_row;
+                if (strip_height > td->td_rowsperstrip)
+                    strip_height = td->td_rowsperstrip;
+                if( TIFFVStripSize64(tif, strip_height) != (uint64)cc )
+                    break;
+            }
+
+            /* Check for overflow */
+            if( (size_t)tif->tif_rawdatasize != (uint64)tif->tif_rawdatasize )
+                break;
+            if( (size_t)cc != (uint64)cc )
+                break;
+
+            /* Go for compression using libdeflate */
+            {
+                size_t nCompressedBytes;
+                if( sp->libdeflate_enc == NULL )
+                {
+                    /* To get results as good as zlib, we asked for an extra */
+                    /* level of compression */
+                    sp->libdeflate_enc = libdeflate_alloc_compressor(
+                        sp->zipquality == Z_DEFAULT_COMPRESSION ? 7 :
+                        sp->zipquality >= 6 && sp->zipquality <= 9 ? sp->zipquality + 1 :
+                        sp->zipquality);
+                    if( sp->libdeflate_enc == NULL )
+                    {
+                        TIFFErrorExt(tif->tif_clientdata, module,
+                                    "Cannot allocate compressor");
+                        break;
+                    }
+                }
+
+                /* Make sure the output buffer is large enough for the worse case. */
+                /* In TIFFWriteBufferSetup(), when libtiff allocates the buffer */
+                /* we've taken a 10% margin over the uncompressed size, which should */
+                /* be large enough even for the the worse case scenario. */
+                if( libdeflate_zlib_compress_bound(sp->libdeflate_enc, (size_t)cc) >
+                        (size_t)tif->tif_rawdatasize)
+                {
+                    break;
+                }
+
+                sp->libdeflate_state = 1;
+                nCompressedBytes = libdeflate_zlib_compress(
+                    sp->libdeflate_enc, bp, (size_t)cc, tif->tif_rawdata, (size_t)tif->tif_rawdatasize);
+
+                if( nCompressedBytes == 0 )
+                {
+                    TIFFErrorExt(tif->tif_clientdata, module,
+                                 "Encoder error at scanline %lu",
+                                 (unsigned long) tif->tif_row);
+                    return 0;
+                }
+
+                tif->tif_rawcc = nCompressedBytes;
+
+                if( !TIFFFlushData1(tif) )
+                    return 0;
+
+                return 1;
+            }
+        } while(0);
+        sp->libdeflate_state = 0;
+#endif /* LIBDEFLATE_SUPPORT */
+
 	sp->stream.next_in = bp;
 	assert(sizeof(sp->stream.avail_in)==4);  /* if this assert gets raised,
 	    we need to simplify this code to reflect a ZLib that is likely updated
@@ -265,7 +440,8 @@ ZIPEncode(TIFF* tif, uint8* bp, tmsize_t cc, uint16 s)
 		}
 		if (sp->stream.avail_out == 0) {
 			tif->tif_rawcc = tif->tif_rawdatasize;
-			TIFFFlushData1(tif);
+			if (!TIFFFlushData1(tif))
+				return 0;
 			sp->stream.next_out = tif->tif_rawdata;
 			sp->stream.avail_out = (uint64)tif->tif_rawdatasize <= 0xFFFFFFFFU ? (uInt)tif->tif_rawdatasize : 0xFFFFFFFFU;
 		}
@@ -285,6 +461,11 @@ ZIPPostEncode(TIFF* tif)
 	ZIPState *sp = EncoderState(tif);
 	int state;

+#if LIBDEFLATE_SUPPORT
+        if( sp->libdeflate_state == 1 )
+            return 1;
+#endif
+
 	sp->stream.avail_in = 0;
 	do {
 		state = deflate(&sp->stream, Z_FINISH);
@@ -294,7 +475,8 @@ ZIPPostEncode(TIFF* tif)
 			if ((tmsize_t)sp->stream.avail_out != tif->tif_rawdatasize)
 			{
 				tif->tif_rawcc =  tif->tif_rawdatasize - sp->stream.avail_out;
-				TIFFFlushData1(tif);
+				if (!TIFFFlushData1(tif))
+					return 0;
 				sp->stream.next_out = tif->tif_rawdata;
 				sp->stream.avail_out = (uint64)tif->tif_rawdatasize <= 0xFFFFFFFFU ? (uInt)tif->tif_rawdatasize : 0xFFFFFFFFU;
 			}
@@ -327,6 +509,14 @@ ZIPCleanup(TIFF* tif)
 		inflateEnd(&sp->stream);
 		sp->state = 0;
 	}
+
+#if LIBDEFLATE_SUPPORT
+        if( sp->libdeflate_dec )
+            libdeflate_free_decompressor(sp->libdeflate_dec);
+        if( sp->libdeflate_enc )
+            libdeflate_free_compressor(sp->libdeflate_enc);
+#endif
+
 	_TIFFfree(sp);
 	tif->tif_data = NULL;

@@ -342,15 +532,55 @@ ZIPVSetField(TIFF* tif, uint32 tag, va_list ap)
 	switch (tag) {
 	case TIFFTAG_ZIPQUALITY:
 		sp->zipquality = (int) va_arg(ap, int);
-		if ( sp->state&ZSTATE_INIT_ENCODE ) {
+                if( sp->zipquality < Z_DEFAULT_COMPRESSION ||
+                    sp->zipquality > LIBDEFLATE_MAX_COMPRESSION_LEVEL ) {
+                    TIFFErrorExt(tif->tif_clientdata, module,
+                                 "Invalid ZipQuality value. Should be in [-1,%d] range",
+                                 LIBDEFLATE_MAX_COMPRESSION_LEVEL);
+                    return 0;
+                }
+
+                if ( sp->state&ZSTATE_INIT_ENCODE ) {
+                        int cappedQuality = sp->zipquality;
+                        if( cappedQuality > Z_BEST_COMPRESSION )
+                            cappedQuality = Z_BEST_COMPRESSION;
 			if (deflateParams(&sp->stream,
-			    sp->zipquality, Z_DEFAULT_STRATEGY) != Z_OK) {
+			    cappedQuality, Z_DEFAULT_STRATEGY) != Z_OK) {
 				TIFFErrorExt(tif->tif_clientdata, module, "ZLib error: %s",
 					     SAFE_MSG(sp));
 				return (0);
 			}
 		}
+
+#if LIBDEFLATE_SUPPORT
+                if( sp->libdeflate_enc )
+                {
+                    libdeflate_free_compressor(sp->libdeflate_enc);
+                    sp->libdeflate_enc = NULL;
+                }
+#endif
+
 		return (1);
+
+        case TIFFTAG_DEFLATE_SUBCODEC:
+                sp->subcodec = (int) va_arg(ap, int);
+                if( sp->subcodec != DEFLATE_SUBCODEC_ZLIB &&
+                    sp->subcodec != DEFLATE_SUBCODEC_LIBDEFLATE )
+                {
+                    TIFFErrorExt(tif->tif_clientdata, module,
+                                 "Invalid DeflateCodec value.");
+                    return 0;
+                }
+#if !LIBDEFLATE_SUPPORT
+                if( sp->subcodec == DEFLATE_SUBCODEC_LIBDEFLATE )
+                {
+                    TIFFErrorExt(tif->tif_clientdata, module,
+                                 "DeflateCodec = DEFLATE_SUBCODEC_LIBDEFLATE unsupported in this build");
+                    return 0;
+                }
+#endif
+                return 1;
+
 	default:
 		return (*sp->vsetparent)(tif, tag, ap);
 	}
@@ -366,6 +596,11 @@ ZIPVGetField(TIFF* tif, uint32 tag, va_list ap)
 	case TIFFTAG_ZIPQUALITY:
 		*va_arg(ap, int*) = sp->zipquality;
 		break;
+
+        case TIFFTAG_DEFLATE_SUBCODEC:
+		*va_arg(ap, int*) = sp->subcodec;
+		break;
+
 	default:
 		return (*sp->vgetparent)(tif, tag, ap);
 	}
@@ -374,6 +609,7 @@ ZIPVGetField(TIFF* tif, uint32 tag, va_list ap)

 static const TIFFField zipFields[] = {
    { TIFFTAG_ZIPQUALITY, 0, 0, TIFF_ANY, 0, TIFF_SETGET_INT, TIFF_SETGET_UNDEFINED, FIELD_PSEUDO, TRUE, FALSE, "", NULL },
+    { TIFFTAG_DEFLATE_SUBCODEC, 0, 0, TIFF_ANY, 0, TIFF_SETGET_INT, TIFF_SETGET_UNDEFINED, FIELD_PSEUDO, TRUE, FALSE, "", NULL },
 };

 int
@@ -384,6 +620,9 @@ TIFFInitZIP(TIFF* tif, int scheme)

 	assert( (scheme == COMPRESSION_DEFLATE)
 		|| (scheme == COMPRESSION_ADOBE_DEFLATE));
+#ifdef NDEBUG
+	(void)scheme;
+#endif

 	/*
 	 * Merge codec-specific tag information.
@@ -397,7 +636,7 @@ TIFFInitZIP(TIFF* tif, int scheme)
 	/*
 	 * Allocate state block so tag methods have storage to record values.
 	 */
-	tif->tif_data = (uint8*) _TIFFmalloc(sizeof (ZIPState));
+	tif->tif_data = (uint8*) _TIFFcalloc(sizeof (ZIPState), 1);
 	if (tif->tif_data == NULL)
 		goto bad;
 	sp = ZState(tif);
@@ -417,6 +656,11 @@ TIFFInitZIP(TIFF* tif, int scheme)
 	/* Default values for codec-specific fields */
 	sp->zipquality = Z_DEFAULT_COMPRESSION;	/* default comp. level */
 	sp->state = 0;
+#if LIBDEFLATE_SUPPORT
+        sp->subcodec = DEFLATE_SUBCODEC_LIBDEFLATE;
+#else
+        sp->subcodec = DEFLATE_SUBCODEC_ZLIB;
+#endif

 	/*
 	 * Install codec methods.
@@ -260,7 +260,8 @@ ZSTDEncode(TIFF* tif, uint8* bp, tmsize_t cc, uint16 s)
                }
                if( sp->out_buffer.pos == sp->out_buffer.size ) {
                        tif->tif_rawcc = tif->tif_rawdatasize;
-                        TIFFFlushData1(tif);
+                        if (!TIFFFlushData1(tif))
+                                return 0;
                        sp->out_buffer.dst = tif->tif_rawcp;
                        sp->out_buffer.pos = 0;
                }
@@ -289,7 +290,8 @@ ZSTDPostEncode(TIFF* tif)
                }
                if( sp->out_buffer.pos > 0 ) {
                        tif->tif_rawcc = sp->out_buffer.pos;
-                        TIFFFlushData1(tif);
+                        if (!TIFFFlushData1(tif))
+                                return 0;
                        sp->out_buffer.dst = tif->tif_rawcp;
                        sp->out_buffer.pos = 0;
                }
@@ -119,6 +119,11 @@ typedef struct {
 * Tag data type information.
 *
 * Note: RATIONALs are the ratio of two 32-bit integer values.
+ *--:
+ * Note2: TIFF_IFD8 data type is used in tiffFields[]-tag definition in order to distinguish the write-handling 
+          of those tags between ClassicTIFF and BigTiff:
+		  For ClassicTIFF libtiff writes a 32-bit value and the TIFF_IFD type-id into the file
+		  For BigTIFF     libtiff writes a 64-bit value and the TIFF_IFD8 type-id into the file
 */
 typedef enum {
 	TIFF_NOTYPE = 0,      /* placeholder */
@@ -375,6 +380,7 @@ typedef enum {
 						   January 2004 */
 #define TIFFTAG_OPIIMAGEID		32781	/* %OPI ImageID
 						   [Adobe TIFF technote] */
+#define	TIFFTAG_TIFFANNOTATIONDATA	32932	/* http://web.archive.org/web/20050309141348/http://www.kofile.com/support%20pro/faqs/annospec.htm */
 /* tags 32952-32956 are private tags registered to Island Graphics */
 #define TIFFTAG_REFPTS			32953	/* image reference points */
 #define TIFFTAG_REGIONTACKPOINT		32954	/* region-xform tack point */
@@ -409,8 +415,23 @@ typedef enum {
 #define TIFFTAG_CFAPATTERN		33422	/* color filter array pattern */
 /* tag 33432 is listed in the 6.0 spec w/ unknown ownership */
 #define	TIFFTAG_COPYRIGHT		33432	/* copyright string */
+/* Tags 33445-33452 are used for GEL fileformat, see
+ * http://research.stowers-institute.org/mcm/efg/ScientificSoftware/Utility/TiffTags/GEL-FileFormat.pdf
+ */
+#define	TIFFTAG_MD_FILETAG		33445	/* http://research.stowers-institute.org/mcm/efg/ScientificSoftware/Utility/TiffTags/GEL-FileFormat.pdf */
+#define	TIFFTAG_MD_SCALEPIXEL	33446	/* http://research.stowers-institute.org/mcm/efg/ScientificSoftware/Utility/TiffTags/GEL-FileFormat.pdf */
+#define	TIFFTAG_MD_COLORTABLE	33447	/* http://research.stowers-institute.org/mcm/efg/ScientificSoftware/Utility/TiffTags/GEL-FileFormat.pdf */
+#define	TIFFTAG_MD_LABNAME	33448	/* http://research.stowers-institute.org/mcm/efg/ScientificSoftware/Utility/TiffTags/GEL-FileFormat.pdf */
+#define	TIFFTAG_MD_SAMPLEINFO	33449	/* http://research.stowers-institute.org/mcm/efg/ScientificSoftware/Utility/TiffTags/GEL-FileFormat.pdf */
+#define	TIFFTAG_MD_PREPDATE	33450	/* http://research.stowers-institute.org/mcm/efg/ScientificSoftware/Utility/TiffTags/GEL-FileFormat.pdf */
+#define	TIFFTAG_MD_PREPTIME	33451	/* http://research.stowers-institute.org/mcm/efg/ScientificSoftware/Utility/TiffTags/GEL-FileFormat.pdf */
+#define	TIFFTAG_MD_FILEUNITS	33452	/* http://research.stowers-institute.org/mcm/efg/ScientificSoftware/Utility/TiffTags/GEL-FileFormat.pdf */
 /* IPTC TAG from RichTIFF specifications */
 #define TIFFTAG_RICHTIFFIPTC		33723
+#define	TIFFTAG_INGR_PACKET_DATA_TAG	33918	/* Intergraph Application specific storage. */
+#define	TIFFTAG_INGR_FLAG_REGISTERS	33919	/* Intergraph Application specific flags. */
+#define	TIFFTAG_IRASB_TRANSORMATION_MATRIX	33920	/* Originally part of Intergraph's GeoTIFF tags, but likely understood by IrasB only. */
+#define	TIFFTAG_MODELTIEPOINTTAG	33922	/* GeoTIFF */
 /* 34016-34029 are reserved for ANSI IT8 TIFF/IT <dkelly@apago.com) */
 #define TIFFTAG_IT8SITE			34016	/* site name */
 #define TIFFTAG_IT8COLORSEQUENCE	34017	/* color seq. [RGB,CMYK,etc] */
@@ -432,6 +453,7 @@ typedef enum {
 #define TIFFTAG_IT8CMYKEQUIVALENT	34032	/* CMYK color equivalents */
 /* tags 34232-34236 are private tags registered to Texas Instruments */
 #define TIFFTAG_FRAMECOUNT              34232   /* Sequence Frame Count */
+#define TIFFTAG_MODELTRANSFORMATIONTAG	34264	/* Used in interchangeable GeoTIFF files */
 /* tag 34377 is private tag registered to Adobe for PhotoShop */
 #define TIFFTAG_PHOTOSHOP		34377 
 /* tags 34665, 34853 and 40965 are documented in EXIF specification */
@@ -451,7 +473,15 @@ typedef enum {
 #define TIFFTAG_STONITS			37439	/* Sample value to Nits */
 /* tag 34929 is a private tag registered to FedEx */
 #define	TIFFTAG_FEDEX_EDR		34929	/* unknown use */
+#define	TIFFTAG_IMAGESOURCEDATA		37724	/* http://justsolve.archiveteam.org/wiki/PSD, http://www.adobe.com/devnet-apps/photoshop/fileformatashtml/ */
 #define TIFFTAG_INTEROPERABILITYIFD	40965	/* Pointer to Interoperability private directory */
+#define	TIFFTAG_GDAL_METADATA		42112	/* Used by the GDAL library */
+#define	TIFFTAG_GDAL_NODATA		42113	/* Used by the GDAL library */
+#define	TIFFTAG_OCE_SCANJOB_DESCRIPTION	50215	/* Used in the Oce scanning process */
+#define	TIFFTAG_OCE_APPLICATION_SELECTOR	50216	/* Used in the Oce scanning process. */
+#define	TIFFTAG_OCE_IDENTIFICATION_NUMBER	50217
+#define	TIFFTAG_OCE_IMAGELOGIC_CHARACTERISTICS	50218
+
 /* tags 50674 to 50677 are reserved for ESRI */
 #define TIFFTAG_LERC_PARAMETERS         50674   /* Stores LERC version and additional compression method */
 /* Adobe Digital Negative (DNG) format tags */
@@ -535,6 +565,17 @@ typedef enum {
 						   into ICC profile space */
 #define TIFFTAG_CURRENTICCPROFILE	50833	/* & */
 #define TIFFTAG_CURRENTPREPROFILEMATRIX	50834	/* & */
+
+#define TIFFTAG_RPCCOEFFICIENT          50844   /* Define by GDAL for geospatial georeferencing through RPC: http://geotiff.maptools.org/rpc_prop.html */
+
+#define	TIFFTAG_ALIAS_LAYER_METADATA	50784	/* Alias Sketchbook Pro layer usage description. */
+
+/* GeoTIFF DGIWG */
+#define TIFFTAG_TIFF_RSID               50908   /* https://www.awaresystems.be/imaging/tiff/tifftags/tiff_rsid.html */
+#define TIFFTAG_GEO_METADATA            50909   /* https://www.awaresystems.be/imaging/tiff/tifftags/geo_metadata.html */
+
+#define TIFFTAG_EXTRACAMERAPROFILES 50933  /* http://wwwimages.adobe.com/www.adobe.com/content/dam/Adobe/en/products/photoshop/pdfs/dng_spec_1.4.0.0.pdf */
+
 /* tag 65535 is an undefined tag used by Eastman Kodak */
 #define TIFFTAG_DCSHUESHIFTVALUES       65535   /* hue shift correction data */

@@ -615,8 +656,11 @@ typedef enum {
 #define     LERC_ADD_COMPRESSION_DEFLATE 1
 #define     LERC_ADD_COMPRESSION_ZSTD    2
 #define TIFFTAG_LERC_MAXZERROR          65567    /* LERC maximum error */
-#define TIFFTAG_WEBP_LEVEL		  65568	/* WebP compression level: WARNING not registered in Adobe-maintained registry */
-#define TIFFTAG_WEBP_LOSSLESS		65569	/* WebP lossless/lossy : WARNING not registered in Adobe-maintained registry */
+#define TIFFTAG_WEBP_LEVEL		  65568	/* WebP compression level */
+#define TIFFTAG_WEBP_LOSSLESS		65569	/* WebP lossless/lossy */
+#define	TIFFTAG_DEFLATE_SUBCODEC	65570	/* ZIP codec: to get/set the sub-codec to use. Will default to libdeflate when available */
+#define     DEFLATE_SUBCODEC_ZLIB       0
+#define     DEFLATE_SUBCODEC_LIBDEFLATE 1

 /*
 * EXIF tags
@@ -626,8 +670,8 @@ typedef enum {
 #define EXIFTAG_EXPOSUREPROGRAM		34850	/* Exposure program */
 #define EXIFTAG_SPECTRALSENSITIVITY	34852	/* Spectral sensitivity */
 #define EXIFTAG_ISOSPEEDRATINGS		34855	/* ISO speed rating */
-#define EXIFTAG_OECF			34856	/* Optoelectric conversion
-						   factor */
+#define EXIFTAG_PHOTOGRAPHICSENSITIVITY	34855 /* Photographic Sensitivity (new name for tag 34855) */
+#define EXIFTAG_OECF			34856	/* Optoelectric conversion factor */
 #define EXIFTAG_EXIFVERSION		36864	/* Exif version */
 #define EXIFTAG_DATETIMEORIGINAL	36867	/* Date and time of original
 						   data generation */
@@ -679,10 +723,71 @@ typedef enum {
 #define EXIFTAG_SHARPNESS		41994	/* Sharpness */
 #define EXIFTAG_DEVICESETTINGDESCRIPTION 41995	/* Device settings description */
 #define EXIFTAG_SUBJECTDISTANCERANGE	41996	/* Subject distance range */
-#define EXIFTAG_GAINCONTROL		41991	/* Gain control */
-#define EXIFTAG_GAINCONTROL		41991	/* Gain control */
 #define EXIFTAG_IMAGEUNIQUEID		42016	/* Unique image ID */

+/*--: New for EXIF-Version 2.32, May 2019 ... */
+#define EXIFTAG_SENSITIVITYTYPE		34864		/* The SensitivityType tag indicates which one of the parameters of ISO12232 is the PhotographicSensitivity tag. */
+#define EXIFTAG_STANDARDOUTPUTSENSITIVITY		34865		/* This tag indicates the standard output sensitivity value of a camera or input device defined in ISO 12232. */
+#define EXIFTAG_RECOMMENDEDEXPOSUREINDEX		34866		/* recommended exposure index */
+#define EXIFTAG_ISOSPEED		34867		/* ISO speed value */
+#define EXIFTAG_ISOSPEEDLATITUDEYYY		34868		/* ISO speed latitude yyy */
+#define EXIFTAG_ISOSPEEDLATITUDEZZZ		34869		/* ISO speed latitude zzz */
+#define EXIFTAG_OFFSETTIME		36880		/* offset from UTC of the time of DateTime tag. */
+#define EXIFTAG_OFFSETTIMEORIGINAL		36881		/* offset from UTC of the time of DateTimeOriginal tag. */
+#define EXIFTAG_OFFSETTIMEDIGITIZED		36882		/* offset from UTC of the time of DateTimeDigitized tag. */
+#define EXIFTAG_TEMPERATURE		37888		/* Temperature as the ambient situation at the shot in dergee Celsius */
+#define EXIFTAG_HUMIDITY		37889		/* Humidity as the ambient situation at the shot in percent */
+#define EXIFTAG_PRESSURE		37890		/* Pressure as the ambient situation at the shot hecto-Pascal (hPa) */
+#define EXIFTAG_WATERDEPTH		37891		/* WaterDepth as the ambient situation at the shot in meter (m) */
+#define EXIFTAG_ACCELERATION		37892		/* Acceleration (a scalar regardless of direction) as the ambient situation at the shot in units of mGal (10-5 m/s^2) */
+#define EXIFTAG_CAMERAELEVATIONANGLE		37893		/* Elevation/depression. angle of the orientation of the camera(imaging optical axis) as the ambient situation at the shot in degree from -180deg to +180deg. */
+#define EXIFTAG_CAMERAOWNERNAME		42032		/* owner of a camera */
+#define EXIFTAG_BODYSERIALNUMBER		42033		/* serial number of the body of the camera */
+#define EXIFTAG_LENSSPECIFICATION		42034		/* minimum focal length (in mm), maximum focal length (in mm), minimum F number in the minimum focal length, and minimum F number in the maximum focal length, */
+#define EXIFTAG_LENSMAKE		42035		/* the lens manufacturer */
+#define EXIFTAG_LENSMODEL		42036		/* the lens model name and model number */
+#define EXIFTAG_LENSSERIALNUMBER		42037		/* the serial number of the interchangeable lens */
+#define EXIFTAG_GAMMA		42240		/* value of coefficient gamma */
+#define EXIFTAG_COMPOSITEIMAGE		42080	/* composite image */
+#define EXIFTAG_SOURCEIMAGENUMBEROFCOMPOSITEIMAGE		42081	/* source image number of composite image */
+#define EXIFTAG_SOURCEEXPOSURETIMESOFCOMPOSITEIMAGE		42082	/* source exposure times of composite image */
+
+/*
+ * EXIF-GPS tags  (Version 2.31, July 2016)
+ */
+#define		GPSTAG_VERSIONID		0			/* 	Indicates the version of GPSInfoIFD.	 */
+#define		GPSTAG_LATITUDEREF		1			/* 	Indicates whether the latitude is north or south latitude.	 */
+#define		GPSTAG_LATITUDE		2			/* 	Indicates the latitude.	 */
+#define		GPSTAG_LONGITUDEREF		3			/* 	Indicates whether the longitude is east or west longitude.	 */
+#define		GPSTAG_LONGITUDE		4			/* 	Indicates the longitude.	 */
+#define		GPSTAG_ALTITUDEREF		5			/* 	Indicates the altitude used as the reference altitude.	 */
+#define		GPSTAG_ALTITUDE		6			/* 	Indicates the altitude based on the reference in GPSAltitudeRef.	 */
+#define		GPSTAG_TIMESTAMP		7			/* 	Indicates the time as UTC (Coordinated Universal Time).	 */
+#define		GPSTAG_SATELLITES		8			/* 	Indicates the GPS satellites used for measurements.	 */
+#define		GPSTAG_STATUS		9			/* 	Indicates the status of the GPS receiver when the image is recorded.	 */
+#define		GPSTAG_MEASUREMODE		10			/* 	Indicates the GPS measurement mode.	 */
+#define		GPSTAG_DOP		11			/* 	Indicates the GPS DOP (data degree of precision).	 */
+#define		GPSTAG_SPEEDREF		12			/* 	Indicates the unit used to express the GPS receiver speed of movement.	 */
+#define		GPSTAG_SPEED		13			/* 	Indicates the speed of GPS receiver movement.	 */
+#define		GPSTAG_TRACKREF		14			/* 	Indicates the reference for giving the direction of GPS receiver movement.	 */
+#define		GPSTAG_TRACK		15			/* 	Indicates the direction of GPS receiver movement.	 */
+#define		GPSTAG_IMGDIRECTIONREF		16			/* 	Indicates the reference for giving the direction of the image when it is captured.	 */
+#define		GPSTAG_IMGDIRECTION		17			/* 	Indicates the direction of the image when it was captured.	 */
+#define		GPSTAG_MAPDATUM		18			/* 	Indicates the geodetic survey data used by the GPS receiver. (e.g. WGS-84)	 */
+#define		GPSTAG_DESTLATITUDEREF		19			/* 	Indicates whether the latitude of the destination point is north or south latitude.	 */
+#define		GPSTAG_DESTLATITUDE		20			/* 	Indicates the latitude of the destination point.	 */
+#define		GPSTAG_DESTLONGITUDEREF		21			/* 	Indicates whether the longitude of the destination point is east or west longitude.	 */
+#define		GPSTAG_DESTLONGITUDE		22			/* 	Indicates the longitude of the destination point.	 */
+#define		GPSTAG_DESTBEARINGREF		23			/* 	Indicates the reference used for giving the bearing to the destination point.	 */
+#define		GPSTAG_DESTBEARING		24			/* 	Indicates the bearing to the destination point.	 */
+#define		GPSTAG_DESTDISTANCEREF		25			/* 	Indicates the unit used to express the distance to the destination point.	 */
+#define		GPSTAG_DESTDISTANCE		26			/* 	Indicates the distance to the destination point.	 */
+#define		GPSTAG_PROCESSINGMETHOD		27			/* 	A character string recording the name of the method used for location finding.	 */
+#define		GPSTAG_AREAINFORMATION		28			/* 	A character string recording the name of the GPS area.	 */
+#define		GPSTAG_DATESTAMP		29			/* 	A character string recording date and time information relative to UTC (Coordinated Universal Time).	 */
+#define		GPSTAG_DIFFERENTIAL		30			/* 	Indicates whether differential correction is applied to the GPS receiver.	 */
+#define		GPSTAG_GPSHPOSITIONINGERROR		31			/* Indicates horizontal positioning errors in meters.		 */
+
 #endif /* _TIFF_ */

 /* vim: set ts=8 sts=8 sw=8 noet: */
@@ -87,6 +87,9 @@
 /* Support Deflate compression */
 #cmakedefine ZIP_SUPPORT 1

+/* Support libdeflate enhanced compression */
+#cmakedefine LIBDEFLATE_SUPPORT 1
+
 /* Support strip chopping (whether or not to convert single-strip uncompressed
   images to mutiple strips of ~8Kb to reduce memory usage) */
 #cmakedefine STRIPCHOP_DEFAULT 1
@@ -261,8 +261,10 @@ typedef struct {
 #define LOGLUV_PUBLIC 1
 #endif

-#if !defined(__GNUC__) && !defined(__attribute__)
-#  define __attribute__(x) /*nothing*/
+#if defined(__GNUC__) || defined(__attribute__)
+#  define TIFF_ATTRIBUTE(x)    __attribute__(x)
+#else
+#  define TIFF_ATTRIBUTE(x) /*nothing*/
 #endif

 #if defined(c_plusplus) || defined(__cplusplus)
@@ -350,6 +352,7 @@ extern int TIFFVGetFieldDefaulted(TIFF* tif, uint32 tag, va_list ap);
 extern int TIFFReadDirectory(TIFF* tif);
 extern int TIFFReadCustomDirectory(TIFF* tif, toff_t diroff, const TIFFFieldArray* infoarray);
 extern int TIFFReadEXIFDirectory(TIFF* tif, toff_t diroff);
+extern int TIFFReadGPSDirectory(TIFF* tif, toff_t diroff);
 extern uint64 TIFFScanlineSize64(TIFF* tif);
 extern tmsize_t TIFFScanlineSize(TIFF* tif);
 extern uint64 TIFFRasterScanlineSize64(TIFF* tif);
@@ -400,6 +403,7 @@ extern void TIFFFreeDirectory(TIFF*);
 extern int TIFFCreateDirectory(TIFF*);
 extern int TIFFCreateCustomDirectory(TIFF*,const TIFFFieldArray*);
 extern int TIFFCreateEXIFDirectory(TIFF*);
+extern int TIFFCreateGPSDirectory(TIFF*);
 extern int TIFFLastDirectory(TIFF*);
 extern int TIFFSetDirectory(TIFF*, uint16);
 extern int TIFFSetSubDirectory(TIFF*, uint64);
@@ -411,6 +415,8 @@ extern int TIFFWriteDirectory(TIFF *);
 extern int TIFFWriteCustomDirectory(TIFF *, uint64 *);
 extern int TIFFCheckpointDirectory(TIFF *);
 extern int TIFFRewriteDirectory(TIFF *);
+extern int TIFFDeferStrileArrayWriting(TIFF *);
+extern int TIFFForceStrileArrayWriting(TIFF* );

 #if defined(c_plusplus) || defined(__cplusplus)
 extern void TIFFPrintDirectory(TIFF*, FILE*, long = 0);
@@ -448,10 +454,10 @@ extern TIFF* TIFFClientOpen(const char*, const char*,
 	    TIFFMapFileProc, TIFFUnmapFileProc);
 extern const char* TIFFFileName(TIFF*);
 extern const char* TIFFSetFileName(TIFF*, const char *);
-extern void TIFFError(const char*, const char*, ...) __attribute__((__format__ (__printf__,2,3)));
-extern void TIFFErrorExt(thandle_t, const char*, const char*, ...) __attribute__((__format__ (__printf__,3,4)));
-extern void TIFFWarning(const char*, const char*, ...) __attribute__((__format__ (__printf__,2,3)));
-extern void TIFFWarningExt(thandle_t, const char*, const char*, ...) __attribute__((__format__ (__printf__,3,4)));
+extern void TIFFError(const char*, const char*, ...) TIFF_ATTRIBUTE((__format__ (__printf__,2,3)));
+extern void TIFFErrorExt(thandle_t, const char*, const char*, ...) TIFF_ATTRIBUTE((__format__ (__printf__,3,4)));
+extern void TIFFWarning(const char*, const char*, ...) TIFF_ATTRIBUTE((__format__ (__printf__,2,3)));
+extern void TIFFWarningExt(thandle_t, const char*, const char*, ...) TIFF_ATTRIBUTE((__format__ (__printf__,3,4)));
 extern TIFFErrorHandler TIFFSetErrorHandler(TIFFErrorHandler);
 extern TIFFErrorHandlerExt TIFFSetErrorHandlerExt(TIFFErrorHandlerExt);
 extern TIFFErrorHandler TIFFSetWarningHandler(TIFFErrorHandler);
@@ -468,6 +474,9 @@ extern tmsize_t TIFFReadEncodedStrip(TIFF* tif, uint32 strip, void* buf, tmsize_
 extern tmsize_t TIFFReadRawStrip(TIFF* tif, uint32 strip, void* buf, tmsize_t size);  
 extern tmsize_t TIFFReadEncodedTile(TIFF* tif, uint32 tile, void* buf, tmsize_t size);  
 extern tmsize_t TIFFReadRawTile(TIFF* tif, uint32 tile, void* buf, tmsize_t size);  
+extern int      TIFFReadFromUserBuffer(TIFF* tif, uint32 strile,
+                                       void* inbuf, tmsize_t insize,
+                                       void* outbuf, tmsize_t outsize);
 extern tmsize_t TIFFWriteEncodedStrip(TIFF* tif, uint32 strip, void* data, tmsize_t cc);
 extern tmsize_t TIFFWriteRawStrip(TIFF* tif, uint32 strip, void* data, tmsize_t cc);  
 extern tmsize_t TIFFWriteEncodedTile(TIFF* tif, uint32 tile, void* data, tmsize_t cc);  
@@ -488,6 +497,11 @@ extern void TIFFSwabArrayOfDouble(double* dp, tmsize_t n);
 extern void TIFFReverseBits(uint8* cp, tmsize_t n);
 extern const unsigned char* TIFFGetBitRevTable(int);

+extern uint64 TIFFGetStrileOffset(TIFF *tif, uint32 strile);
+extern uint64 TIFFGetStrileByteCount(TIFF *tif, uint32 strile);
+extern uint64 TIFFGetStrileOffsetWithErr(TIFF *tif, uint32 strile, int *pbErr);
+extern uint64 TIFFGetStrileByteCountWithErr(TIFF *tif, uint32 strile, int *pbErr);
+
 #ifdef LOGLUV_PUBLIC
 #define U_NEU		0.210526316
 #define V_NEU		0.473684211
@@ -77,6 +77,19 @@ extern int snprintf(char* str, size_t size, const char* format, ...);
 #define	FALSE	0
 #endif

+#define TIFF_SIZE_T_MAX ((size_t) ~ ((size_t)0))
+#define TIFF_TMSIZE_T_MAX (tmsize_t)(TIFF_SIZE_T_MAX >> 1)
+
+/*
+ * Largest 32-bit unsigned integer value.
+ */
+#define TIFF_UINT32_MAX 0xFFFFFFFFU
+
+/*
+ * Largest 64-bit unsigned integer value.
+ */
+#define TIFF_UINT64_MAX (((uint64)(TIFF_UINT32_MAX)) << 32 | TIFF_UINT32_MAX)
+
 typedef struct client_info {
    struct client_info *next;
    void *data;
@@ -127,6 +140,9 @@ struct tiff {
        #define TIFF_DIRTYSTRIP 0x200000U /* stripoffsets/stripbytecount dirty*/
        #define TIFF_PERSAMPLE  0x400000U /* get/set per sample tags as arrays */
        #define TIFF_BUFFERMMAP 0x800000U /* read buffer (tif_rawdata) points into mmap() memory */
+        #define TIFF_DEFERSTRILELOAD 0x1000000U /* defer strip/tile offset/bytecount array loading. */
+        #define TIFF_LAZYSTRILELOAD  0x2000000U /* lazy/ondemand loading of strip/tile offset/bytecount values. Only used if TIFF_DEFERSTRILELOAD is set and in read-only mode */
+        #define TIFF_CHOPPEDUPARRAYS 0x4000000U /* set when allocChoppedUpStripArrays() has modified strip array */
 	uint64               tif_diroff;       /* file offset of current directory */
 	uint64               tif_nextdiroff;   /* file offset of following directory */
 	uint64*              tif_dirlist;      /* list of offsets to already seen directories to prevent IFD looping */
@@ -258,7 +274,7 @@ struct tiff {
 #define TIFFhowmany8_64(x) (((x)&0x07)?((uint64)(x)>>3)+1:(uint64)(x)>>3)
 #define TIFFroundup_64(x, y) (TIFFhowmany_64(x,y)*(y))

-/* Safe multiply which returns zero if there is an integer overflow */
+/* Safe multiply which returns zero if there is an *unsigned* integer overflow. This macro is not safe for *signed* integer types */
 #define TIFFSafeMultiply(t,v,m) ((((t)(m) != (t)0) && (((t)(((v)*(m))/(m))) == (t)(v))) ? (t)((v)*(m)) : (t)0)

 #define TIFFmax(A,B) ((A)>(B)?(A):(B))
@@ -351,6 +367,9 @@ extern uint32 _TIFFDefaultStripSize(TIFF* tif, uint32 s);
 extern void _TIFFDefaultTileSize(TIFF* tif, uint32* tw, uint32* th);
 extern int _TIFFDataSize(TIFFDataType type);

+/*--: Rational2Double: Return size of TIFFSetGetFieldType in bytes. */
+extern int _TIFFSetGetFieldSize(TIFFSetGetFieldType setgettype);
+
 extern void _TIFFsetByteArray(void**, void*, uint32);
 extern void _TIFFsetString(char**, char*);
 extern void _TIFFsetShortArray(uint16**, uint16*, uint32);
@@ -368,6 +387,8 @@ extern TIFFErrorHandlerExt _TIFFerrorHandlerExt;

 extern uint32 _TIFFMultiply32(TIFF*, uint32, uint32, const char*);
 extern uint64 _TIFFMultiply64(TIFF*, uint64, uint64, const char*);
+extern tmsize_t _TIFFMultiplySSize(TIFF*, tmsize_t, tmsize_t, const char*);
+extern tmsize_t _TIFFCastUInt64ToSSize(TIFF*, uint64, const char*);
 extern void* _TIFFCheckMalloc(TIFF*, tmsize_t, tmsize_t, const char*);
 extern void* _TIFFCheckRealloc(TIFF*, void*, tmsize_t, tmsize_t, const char*);

@@ -1,4 +1,4 @@
-#define TIFFLIB_VERSION_STR "LIBTIFF, Version 4.0.10\nCopyright (c) 1988-1996 Sam Leffler\nCopyright (c) 1991-1996 Silicon Graphics, Inc."
+#define TIFFLIB_VERSION_STR "LIBTIFF, Version 4.2.0\nCopyright (c) 1988-1996 Sam Leffler\nCopyright (c) 1991-1996 Silicon Graphics, Inc."
 /*
 * This define can be used in code that requires
 * compilation-related definitions specific to a
@@ -6,4 +6,4 @@
 * version checking should be done based on the
 * string returned by TIFFGetVersion.
 */
-#define TIFFLIB_VERSION 20181110
+#define TIFFLIB_VERSION 20201219
@@ -25,21 +25,16 @@
 static int EmitYUV(const VP8Io* const io, WebPDecParams* const p) {
  WebPDecBuffer* output = p->output;
  const WebPYUVABuffer* const buf = &output->u.YUVA;
-  uint8_t* const y_dst = buf->y + io->mb_y * buf->y_stride;
-  uint8_t* const u_dst = buf->u + (io->mb_y >> 1) * buf->u_stride;
-  uint8_t* const v_dst = buf->v + (io->mb_y >> 1) * buf->v_stride;
+  uint8_t* const y_dst = buf->y + (size_t)io->mb_y * buf->y_stride;
+  uint8_t* const u_dst = buf->u + (size_t)(io->mb_y >> 1) * buf->u_stride;
+  uint8_t* const v_dst = buf->v + (size_t)(io->mb_y >> 1) * buf->v_stride;
  const int mb_w = io->mb_w;
  const int mb_h = io->mb_h;
  const int uv_w = (mb_w + 1) / 2;
  const int uv_h = (mb_h + 1) / 2;
-  int j;
-  for (j = 0; j < mb_h; ++j) {
-    memcpy(y_dst + j * buf->y_stride, io->y + j * io->y_stride, mb_w);
-  }
-  for (j = 0; j < uv_h; ++j) {
-    memcpy(u_dst + j * buf->u_stride, io->u + j * io->uv_stride, uv_w);
-    memcpy(v_dst + j * buf->v_stride, io->v + j * io->uv_stride, uv_w);
-  }
+  WebPCopyPlane(io->y, io->y_stride, y_dst, buf->y_stride, mb_w, mb_h);
+  WebPCopyPlane(io->u, io->uv_stride, u_dst, buf->u_stride, uv_w, uv_h);
+  WebPCopyPlane(io->v, io->uv_stride, v_dst, buf->v_stride, uv_w, uv_h);
  return io->mb_h;
 }

@@ -47,7 +42,7 @@ static int EmitYUV(const VP8Io* const io, WebPDecParams* const p) {
 static int EmitSampledRGB(const VP8Io* const io, WebPDecParams* const p) {
  WebPDecBuffer* const output = p->output;
  WebPRGBABuffer* const buf = &output->u.RGBA;
-  uint8_t* const dst = buf->rgba + io->mb_y * buf->stride;
+  uint8_t* const dst = buf->rgba + (size_t)io->mb_y * buf->stride;
  WebPSamplerProcessPlane(io->y, io->y_stride,
                          io->u, io->v, io->uv_stride,
                          dst, buf->stride, io->mb_w, io->mb_h,
@@ -62,7 +57,7 @@ static int EmitSampledRGB(const VP8Io* const io, WebPDecParams* const p) {
 static int EmitFancyRGB(const VP8Io* const io, WebPDecParams* const p) {
  int num_lines_out = io->mb_h;   // a priori guess
  const WebPRGBABuffer* const buf = &p->output->u.RGBA;
-  uint8_t* dst = buf->rgba + io->mb_y * buf->stride;
+  uint8_t* dst = buf->rgba + (size_t)io->mb_y * buf->stride;
  WebPUpsampleLinePairFunc upsample = WebPUpsamplers[p->output->colorspace];
  const uint8_t* cur_y = io->y;
  const uint8_t* cur_u = io->u;
@@ -133,7 +128,7 @@ static int EmitAlphaYUV(const VP8Io* const io, WebPDecParams* const p,
  const WebPYUVABuffer* const buf = &p->output->u.YUVA;
  const int mb_w = io->mb_w;
  const int mb_h = io->mb_h;
-  uint8_t* dst = buf->a + io->mb_y * buf->a_stride;
+  uint8_t* dst = buf->a + (size_t)io->mb_y * buf->a_stride;
  int j;
  (void)expected_num_lines_out;
  assert(expected_num_lines_out == mb_h);
@@ -186,7 +181,7 @@ static int EmitAlphaRGB(const VP8Io* const io, WebPDecParams* const p,
        (colorspace == MODE_ARGB || colorspace == MODE_Argb);
    const WebPRGBABuffer* const buf = &p->output->u.RGBA;
    int num_rows;
-    const int start_y = GetAlphaSourceRow(io, &alpha, &num_rows);
+    const size_t start_y = GetAlphaSourceRow(io, &alpha, &num_rows);
    uint8_t* const base_rgba = buf->rgba + start_y * buf->stride;
    uint8_t* const dst = base_rgba + (alpha_first ? 0 : 3);
    const int has_alpha = WebPDispatchAlpha(alpha, io->width, mb_w,
@@ -210,7 +205,7 @@ static int EmitAlphaRGBA4444(const VP8Io* const io, WebPDecParams* const p,
    const WEBP_CSP_MODE colorspace = p->output->colorspace;
    const WebPRGBABuffer* const buf = &p->output->u.RGBA;
    int num_rows;
-    const int start_y = GetAlphaSourceRow(io, &alpha, &num_rows);
+    const size_t start_y = GetAlphaSourceRow(io, &alpha, &num_rows);
    uint8_t* const base_rgba = buf->rgba + start_y * buf->stride;
 #if (WEBP_SWAP_16BIT_CSP == 1)
    uint8_t* alpha_dst = base_rgba;
@@ -276,9 +271,9 @@ static int EmitRescaledYUV(const VP8Io* const io, WebPDecParams* const p) {
 static int EmitRescaledAlphaYUV(const VP8Io* const io, WebPDecParams* const p,
                                int expected_num_lines_out) {
  const WebPYUVABuffer* const buf = &p->output->u.YUVA;
-  uint8_t* const dst_a = buf->a + p->last_y * buf->a_stride;
+  uint8_t* const dst_a = buf->a + (size_t)p->last_y * buf->a_stride;
  if (io->a != NULL) {
-    uint8_t* const dst_y = buf->y + p->last_y * buf->y_stride;
+    uint8_t* const dst_y = buf->y + (size_t)p->last_y * buf->y_stride;
    const int num_lines_out = Rescale(io->a, io->width, io->mb_h, p->scaler_a);
    assert(expected_num_lines_out == num_lines_out);
    if (num_lines_out > 0) {   // unmultiply the Y
@@ -356,7 +351,7 @@ static int ExportRGB(WebPDecParams* const p, int y_pos) {
  const WebPYUV444Converter convert =
      WebPYUV444Converters[p->output->colorspace];
  const WebPRGBABuffer* const buf = &p->output->u.RGBA;
-  uint8_t* dst = buf->rgba + y_pos * buf->stride;
+  uint8_t* dst = buf->rgba + (size_t)y_pos * buf->stride;
  int num_lines_out = 0;
  // For RGB rescaling, because of the YUV420, current scan position
  // U/V can be +1/-1 line from the Y one.  Hence the double test.
@@ -383,15 +378,15 @@ static int EmitRescaledRGB(const VP8Io* const io, WebPDecParams* const p) {
  while (j < mb_h) {
    const int y_lines_in =
        WebPRescalerImport(p->scaler_y, mb_h - j,
-                           io->y + j * io->y_stride, io->y_stride);
+                           io->y + (size_t)j * io->y_stride, io->y_stride);
    j += y_lines_in;
    if (WebPRescaleNeededLines(p->scaler_u, uv_mb_h - uv_j)) {
-      const int u_lines_in =
-          WebPRescalerImport(p->scaler_u, uv_mb_h - uv_j,
-                             io->u + uv_j * io->uv_stride, io->uv_stride);
-      const int v_lines_in =
-          WebPRescalerImport(p->scaler_v, uv_mb_h - uv_j,
-                             io->v + uv_j * io->uv_stride, io->uv_stride);
+      const int u_lines_in = WebPRescalerImport(
+          p->scaler_u, uv_mb_h - uv_j, io->u + (size_t)uv_j * io->uv_stride,
+          io->uv_stride);
+      const int v_lines_in = WebPRescalerImport(
+          p->scaler_v, uv_mb_h - uv_j, io->v + (size_t)uv_j * io->uv_stride,
+          io->uv_stride);
      (void)v_lines_in;   // remove a gcc warning
      assert(u_lines_in == v_lines_in);
      uv_j += u_lines_in;
@@ -403,7 +398,7 @@ static int EmitRescaledRGB(const VP8Io* const io, WebPDecParams* const p) {

 static int ExportAlpha(WebPDecParams* const p, int y_pos, int max_lines_out) {
  const WebPRGBABuffer* const buf = &p->output->u.RGBA;
-  uint8_t* const base_rgba = buf->rgba + y_pos * buf->stride;
+  uint8_t* const base_rgba = buf->rgba + (size_t)y_pos * buf->stride;
  const WEBP_CSP_MODE colorspace = p->output->colorspace;
  const int alpha_first =
      (colorspace == MODE_ARGB || colorspace == MODE_Argb);
@@ -431,7 +426,7 @@ static int ExportAlpha(WebPDecParams* const p, int y_pos, int max_lines_out) {
 static int ExportAlphaRGBA4444(WebPDecParams* const p, int y_pos,
                               int max_lines_out) {
  const WebPRGBABuffer* const buf = &p->output->u.RGBA;
-  uint8_t* const base_rgba = buf->rgba + y_pos * buf->stride;
+  uint8_t* const base_rgba = buf->rgba + (size_t)y_pos * buf->stride;
 #if (WEBP_SWAP_16BIT_CSP == 1)
  uint8_t* alpha_dst = base_rgba;
 #else
@@ -470,7 +465,7 @@ static int EmitRescaledAlphaRGB(const VP8Io* const io, WebPDecParams* const p,
    int lines_left = expected_num_out_lines;
    const int y_end = p->last_y + lines_left;
    while (lines_left > 0) {
-      const int row_offset = scaler->src_y - io->mb_y;
+      const int64_t row_offset = (int64_t)scaler->src_y - io->mb_y;
      WebPRescalerImport(scaler, io->mb_h + io->mb_y - scaler->src_y,
                         io->a + row_offset * io->width, io->width);
      lines_left -= p->emit_alpha_row(p, y_end - lines_left, lines_left);
@@ -494,13 +494,11 @@ static int GetCoeffsAlt(VP8BitReader* const br,
  return 16;
 }

-static WEBP_TSAN_IGNORE_FUNCTION void InitGetCoeffs(void) {
-  if (GetCoeffs == NULL) {
-    if (VP8GetCPUInfo != NULL && VP8GetCPUInfo(kSlowSSSE3)) {
-      GetCoeffs = GetCoeffsAlt;
-    } else {
-      GetCoeffs = GetCoeffsFast;
-    }
+WEBP_DSP_INIT_FUNC(InitGetCoeffs) {
+  if (VP8GetCPUInfo != NULL && VP8GetCPUInfo(kSlowSSSE3)) {
+    GetCoeffs = GetCoeffsAlt;
+  } else {
+    GetCoeffs = GetCoeffsFast;
  }
 }

@@ -31,7 +31,7 @@ extern "C" {

 // version numbers
 #define DEC_MAJ_VERSION 1
-#define DEC_MIN_VERSION 1
+#define DEC_MIN_VERSION 2
 #define DEC_REV_VERSION 0

 // YUV-cache parameters. Cache is 32-bytes wide (= one cacheline).
@@ -947,7 +947,6 @@ static WEBP_INLINE void CopyBlock8b(uint8_t* const dst, int dist, int length) {
        break;
      default:
        goto Copy;
-        break;
    }
    CopySmallPattern8b(src, dst, length, pattern);
    return;
@@ -346,12 +346,15 @@ int WebPAnimDecoderGetNext(WebPAnimDecoder* dec,
  {
    const uint8_t* in = iter.fragment.bytes;
    const size_t in_size = iter.fragment.size;
-    const size_t out_offset =
-        (iter.y_offset * width + iter.x_offset) * NUM_CHANNELS;
+    const uint32_t stride = width * NUM_CHANNELS;  // at most 25 + 2 bits
+    const uint64_t out_offset = (uint64_t)iter.y_offset * stride +
+                                (uint64_t)iter.x_offset * NUM_CHANNELS;  // 53b
+    const uint64_t size = (uint64_t)iter.height * stride;  // at most 25 + 27b
    WebPDecoderConfig* const config = &dec->config_;
    WebPRGBABuffer* const buf = &config->output.u.RGBA;
-    buf->stride = NUM_CHANNELS * width;
-    buf->size = buf->stride * iter.height;
+    if ((size_t)size != size) goto Error;
+    buf->stride = (int)stride;
+    buf->size = (size_t)size;
    buf->rgba = dec->curr_frame_ + out_offset;

    if (WebPDecode(in, in_size, config) != VP8_STATUS_OK) {
@@ -24,7 +24,7 @@
 #include "src/webp/format_constants.h"

 #define DMUX_MAJ_VERSION 1
-#define DMUX_MIN_VERSION 1
+#define DMUX_MIN_VERSION 2
 #define DMUX_REV_VERSION 0

 typedef struct {
@@ -312,6 +312,7 @@ static ParseStatus ParseAnimationFrame(
  int bits;
  MemBuffer* const mem = &dmux->mem_;
  Frame* frame;
+  size_t start_offset;
  ParseStatus status =
      NewFrame(mem, ANMF_CHUNK_SIZE, frame_chunk_size, &frame);
  if (status != PARSE_OK) return status;
@@ -332,7 +333,11 @@ static ParseStatus ParseAnimationFrame(

  // Store a frame only if the animation flag is set there is some data for
  // this frame is available.
+  start_offset = mem->start_;
  status = StoreFrame(dmux->num_frames_ + 1, anmf_payload_size, mem, frame);
+  if (status != PARSE_ERROR && mem->start_ - start_offset > anmf_payload_size) {
+    status = PARSE_ERROR;
+  }
  if (status != PARSE_ERROR && is_animation && frame->frame_num_ > 0) {
    added_frame = AddFrame(dmux, frame);
    if (added_frame) {
@@ -359,6 +359,11 @@ static int HasAlpha32b_C(const uint8_t* src, int length) {
  return 0;
 }

+static void AlphaReplace_C(uint32_t* src, int length, uint32_t color) {
+  int x;
+  for (x = 0; x < length; ++x) if ((src[x] >> 24) == 0) src[x] = color;
+}
+
 //------------------------------------------------------------------------------
 // Simple channel manipulations.

@@ -400,6 +405,7 @@ void (*WebPPackRGB)(const uint8_t* r, const uint8_t* g, const uint8_t* b,

 int (*WebPHasAlpha8b)(const uint8_t* src, int length);
 int (*WebPHasAlpha32b)(const uint8_t* src, int length);
+void (*WebPAlphaReplace)(uint32_t* src, int length, uint32_t color);

 //------------------------------------------------------------------------------
 // Init function
@@ -428,6 +434,7 @@ WEBP_DSP_INIT_FUNC(WebPInitAlphaProcessing) {

  WebPHasAlpha8b = HasAlpha8b_C;
  WebPHasAlpha32b = HasAlpha32b_C;
+  WebPAlphaReplace = AlphaReplace_C;

  // If defined, use CPUInfo() to overwrite some pointers with faster versions.
  if (VP8GetCPUInfo != NULL) {
@@ -469,4 +476,5 @@ WEBP_DSP_INIT_FUNC(WebPInitAlphaProcessing) {
  assert(WebPPackRGB != NULL);
  assert(WebPHasAlpha8b != NULL);
  assert(WebPHasAlpha32b != NULL);
+  assert(WebPAlphaReplace != NULL);
 }
@@ -265,6 +265,27 @@ static int HasAlpha32b_SSE2(const uint8_t* src, int length) {
  return 0;
 }

+static void AlphaReplace_SSE2(uint32_t* src, int length, uint32_t color) {
+  const __m128i m_color = _mm_set1_epi32(color);
+  const __m128i zero = _mm_setzero_si128();
+  int i = 0;
+  for (; i + 8 <= length; i += 8) {
+    const __m128i a0 = _mm_loadu_si128((const __m128i*)(src + i + 0));
+    const __m128i a1 = _mm_loadu_si128((const __m128i*)(src + i + 4));
+    const __m128i b0 = _mm_srai_epi32(a0, 24);
+    const __m128i b1 = _mm_srai_epi32(a1, 24);
+    const __m128i c0 = _mm_cmpeq_epi32(b0, zero);
+    const __m128i c1 = _mm_cmpeq_epi32(b1, zero);
+    const __m128i d0 = _mm_and_si128(c0, m_color);
+    const __m128i d1 = _mm_and_si128(c1, m_color);
+    const __m128i e0 = _mm_andnot_si128(c0, a0);
+    const __m128i e1 = _mm_andnot_si128(c1, a1);
+    _mm_storeu_si128((__m128i*)(src + i + 0), _mm_or_si128(d0, e0));
+    _mm_storeu_si128((__m128i*)(src + i + 4), _mm_or_si128(d1, e1));
+  }
+  for (; i < length; ++i) if ((src[i] >> 24) == 0) src[i] = color;
+}
+
 // -----------------------------------------------------------------------------
 // Apply alpha value to rows

@@ -334,6 +355,7 @@ WEBP_TSAN_IGNORE_FUNCTION void WebPInitAlphaProcessingSSE2(void) {

  WebPHasAlpha8b = HasAlpha8b_SSE2;
  WebPHasAlpha32b = HasAlpha32b_SSE2;
+  WebPAlphaReplace = AlphaReplace_SSE2;
 }

 #else  // !WEBP_USE_SSE2
@@ -55,12 +55,18 @@ static WEBP_INLINE void GetCPUInfo(int cpu_info[4], int info_type) {
    : "=a"(cpu_info[0]), "=b"(cpu_info[1]), "=c"(cpu_info[2]), "=d"(cpu_info[3])
    : "a"(info_type), "c"(0));
 }
-#elif (defined(_M_X64) || defined(_M_IX86)) && \
-      defined(_MSC_FULL_VER) && _MSC_FULL_VER >= 150030729  // >= VS2008 SP1
+#elif defined(_MSC_VER) && (defined(_M_X64) || defined(_M_IX86))
+
+#if defined(_MSC_FULL_VER) && _MSC_FULL_VER >= 150030729  // >= VS2008 SP1
 #include <intrin.h>
 #define GetCPUInfo(info, type) __cpuidex(info, type, 0)  // set ecx=0
-#elif defined(WEBP_MSC_SSE2)
+#define WEBP_HAVE_MSC_CPUID
+#elif _MSC_VER > 1310
+#include <intrin.h>
 #define GetCPUInfo __cpuid
+#define WEBP_HAVE_MSC_CPUID
+#endif
+
 #endif

 // NaCl has no support for xgetbv or the raw opcode.
@@ -94,7 +100,7 @@ static WEBP_INLINE uint64_t xgetbv(void) {
 #define xgetbv() 0U  // no AVX for older x64 or unrecognized toolchains.
 #endif

-#if defined(__i386__) || defined(__x86_64__) || defined(WEBP_MSC_SSE2)
+#if defined(__i386__) || defined(__x86_64__) || defined(WEBP_HAVE_MSC_CPUID)

 // helper function for run-time detection of slow SSSE3 platforms
 static int CheckSlowModel(int info) {
@@ -179,6 +185,30 @@ static int AndroidCPUInfo(CPUFeature feature) {
  return 0;
 }
 VP8CPUInfo VP8GetCPUInfo = AndroidCPUInfo;
+#elif defined(EMSCRIPTEN) // also needs to be before generic NEON test
+// Use compile flags as an indicator of SIMD support instead of a runtime check.
+static int wasmCPUInfo(CPUFeature feature) {
+  switch (feature) {
+#ifdef WEBP_USE_SSE2
+    case kSSE2:
+      return 1;
+#endif
+#ifdef WEBP_USE_SSE41
+    case kSSE3:
+    case kSlowSSSE3:
+    case kSSE4_1:
+      return 1;
+#endif
+#ifdef WEBP_USE_NEON
+    case kNEON:
+      return 1;
+#endif
+    default:
+      break;
+  }
+  return 0;
+}
+VP8CPUInfo VP8GetCPUInfo = wasmCPUInfo;
 #elif defined(WEBP_USE_NEON)
 // define a dummy function to enable turning off NEON at runtime by setting
 // VP8DecGetCPUInfo = NULL
@@ -1283,12 +1283,12 @@ static void DC4_NEON(uint8_t* dst) {    // DC
  const uint8x8_t A = vld1_u8(dst - BPS);  // top row
  const uint16x4_t p0 = vpaddl_u8(A);  // cascading summation of the top
  const uint16x4_t p1 = vpadd_u16(p0, p0);
-  const uint16x8_t L0 = vmovl_u8(vld1_u8(dst + 0 * BPS - 1));
-  const uint16x8_t L1 = vmovl_u8(vld1_u8(dst + 1 * BPS - 1));
-  const uint16x8_t L2 = vmovl_u8(vld1_u8(dst + 2 * BPS - 1));
-  const uint16x8_t L3 = vmovl_u8(vld1_u8(dst + 3 * BPS - 1));
-  const uint16x8_t s0 = vaddq_u16(L0, L1);
-  const uint16x8_t s1 = vaddq_u16(L2, L3);
+  const uint8x8_t L0 = vld1_u8(dst + 0 * BPS - 1);
+  const uint8x8_t L1 = vld1_u8(dst + 1 * BPS - 1);
+  const uint8x8_t L2 = vld1_u8(dst + 2 * BPS - 1);
+  const uint8x8_t L3 = vld1_u8(dst + 3 * BPS - 1);
+  const uint16x8_t s0 = vaddl_u8(L0, L1);
+  const uint16x8_t s1 = vaddl_u8(L2, L3);
  const uint16x8_t s01 = vaddq_u16(s0, s1);
  const uint16x8_t sum = vaddq_u16(s01, vcombine_u16(p1, p1));
  const uint8x8_t dc0 = vrshrn_n_u16(sum, 3);  // (sum + 4) >> 3
@@ -1429,8 +1429,7 @@ static WEBP_INLINE void DC8_NEON(uint8_t* dst, int do_top, int do_left) {
  if (do_top) {
    const uint8x8_t A = vld1_u8(dst - BPS);  // top row
 #if defined(__aarch64__)
-    const uint16x8_t B = vmovl_u8(A);
-    const uint16_t p2 = vaddvq_u16(B);
+    const uint16_t p2 = vaddlv_u8(A);
    sum_top = vdupq_n_u16(p2);
 #else
    const uint16x4_t p0 = vpaddl_u8(A);  // cascading summation of the top
@@ -1441,18 +1440,18 @@ static WEBP_INLINE void DC8_NEON(uint8_t* dst, int do_top, int do_left) {
  }

  if (do_left) {
-    const uint16x8_t L0 = vmovl_u8(vld1_u8(dst + 0 * BPS - 1));
-    const uint16x8_t L1 = vmovl_u8(vld1_u8(dst + 1 * BPS - 1));
-    const uint16x8_t L2 = vmovl_u8(vld1_u8(dst + 2 * BPS - 1));
-    const uint16x8_t L3 = vmovl_u8(vld1_u8(dst + 3 * BPS - 1));
-    const uint16x8_t L4 = vmovl_u8(vld1_u8(dst + 4 * BPS - 1));
-    const uint16x8_t L5 = vmovl_u8(vld1_u8(dst + 5 * BPS - 1));
-    const uint16x8_t L6 = vmovl_u8(vld1_u8(dst + 6 * BPS - 1));
-    const uint16x8_t L7 = vmovl_u8(vld1_u8(dst + 7 * BPS - 1));
-    const uint16x8_t s0 = vaddq_u16(L0, L1);
-    const uint16x8_t s1 = vaddq_u16(L2, L3);
-    const uint16x8_t s2 = vaddq_u16(L4, L5);
-    const uint16x8_t s3 = vaddq_u16(L6, L7);
+    const uint8x8_t L0 = vld1_u8(dst + 0 * BPS - 1);
+    const uint8x8_t L1 = vld1_u8(dst + 1 * BPS - 1);
+    const uint8x8_t L2 = vld1_u8(dst + 2 * BPS - 1);
+    const uint8x8_t L3 = vld1_u8(dst + 3 * BPS - 1);
+    const uint8x8_t L4 = vld1_u8(dst + 4 * BPS - 1);
+    const uint8x8_t L5 = vld1_u8(dst + 5 * BPS - 1);
+    const uint8x8_t L6 = vld1_u8(dst + 6 * BPS - 1);
+    const uint8x8_t L7 = vld1_u8(dst + 7 * BPS - 1);
+    const uint16x8_t s0 = vaddl_u8(L0, L1);
+    const uint16x8_t s1 = vaddl_u8(L2, L3);
+    const uint16x8_t s2 = vaddl_u8(L4, L5);
+    const uint16x8_t s3 = vaddl_u8(L6, L7);
    const uint16x8_t s01 = vaddq_u16(s0, s1);
    const uint16x8_t s23 = vaddq_u16(s2, s3);
    sum_left = vaddq_u16(s01, s23);
@@ -1512,29 +1511,34 @@ static WEBP_INLINE void DC16_NEON(uint8_t* dst, int do_top, int do_left) {

  if (do_top) {
    const uint8x16_t A = vld1q_u8(dst - BPS);  // top row
+#if defined(__aarch64__)
+    const uint16_t p3 = vaddlvq_u8(A);
+    sum_top = vdupq_n_u16(p3);
+#else
    const uint16x8_t p0 = vpaddlq_u8(A);  // cascading summation of the top
    const uint16x4_t p1 = vadd_u16(vget_low_u16(p0), vget_high_u16(p0));
    const uint16x4_t p2 = vpadd_u16(p1, p1);
    const uint16x4_t p3 = vpadd_u16(p2, p2);
    sum_top = vcombine_u16(p3, p3);
+#endif
  }

  if (do_left) {
    int i;
    sum_left = vdupq_n_u16(0);
    for (i = 0; i < 16; i += 8) {
-      const uint16x8_t L0 = vmovl_u8(vld1_u8(dst + (i + 0) * BPS - 1));
-      const uint16x8_t L1 = vmovl_u8(vld1_u8(dst + (i + 1) * BPS - 1));
-      const uint16x8_t L2 = vmovl_u8(vld1_u8(dst + (i + 2) * BPS - 1));
-      const uint16x8_t L3 = vmovl_u8(vld1_u8(dst + (i + 3) * BPS - 1));
-      const uint16x8_t L4 = vmovl_u8(vld1_u8(dst + (i + 4) * BPS - 1));
-      const uint16x8_t L5 = vmovl_u8(vld1_u8(dst + (i + 5) * BPS - 1));
-      const uint16x8_t L6 = vmovl_u8(vld1_u8(dst + (i + 6) * BPS - 1));
-      const uint16x8_t L7 = vmovl_u8(vld1_u8(dst + (i + 7) * BPS - 1));
-      const uint16x8_t s0 = vaddq_u16(L0, L1);
-      const uint16x8_t s1 = vaddq_u16(L2, L3);
-      const uint16x8_t s2 = vaddq_u16(L4, L5);
-      const uint16x8_t s3 = vaddq_u16(L6, L7);
+      const uint8x8_t L0 = vld1_u8(dst + (i + 0) * BPS - 1);
+      const uint8x8_t L1 = vld1_u8(dst + (i + 1) * BPS - 1);
+      const uint8x8_t L2 = vld1_u8(dst + (i + 2) * BPS - 1);
+      const uint8x8_t L3 = vld1_u8(dst + (i + 3) * BPS - 1);
+      const uint8x8_t L4 = vld1_u8(dst + (i + 4) * BPS - 1);
+      const uint8x8_t L5 = vld1_u8(dst + (i + 5) * BPS - 1);
+      const uint8x8_t L6 = vld1_u8(dst + (i + 6) * BPS - 1);
+      const uint8x8_t L7 = vld1_u8(dst + (i + 7) * BPS - 1);
+      const uint16x8_t s0 = vaddl_u8(L0, L1);
+      const uint16x8_t s1 = vaddl_u8(L2, L3);
+      const uint16x8_t s2 = vaddl_u8(L4, L5);
+      const uint16x8_t s3 = vaddl_u8(L6, L7);
      const uint16x8_t s01 = vaddq_u16(s0, s1);
      const uint16x8_t s23 = vaddq_u16(s2, s3);
      const uint16x8_t sum = vaddq_u16(s01, s23);
@@ -51,9 +51,7 @@ extern "C" {
 # define __has_builtin(x) 0
 #endif

-// for now, none of the optimizations below are available in emscripten
-#if !defined(EMSCRIPTEN)
-
+#if !defined(HAVE_CONFIG_H)
 #if defined(_MSC_VER) && _MSC_VER > 1310 && \
    (defined(_M_X64) || defined(_M_IX86))
 #define WEBP_MSC_SSE2  // Visual C++ SSE2 targets
@@ -63,6 +61,7 @@ extern "C" {
    (defined(_M_X64) || defined(_M_IX86))
 #define WEBP_MSC_SSE41  // Visual C++ SSE4.1 targets
 #endif
+#endif

 // WEBP_HAVE_* are used to indicate the presence of the instruction set in dsp
 // files without intrinsics, allowing the corresponding Init() to be called.
@@ -76,6 +75,9 @@ extern "C" {
 #define WEBP_USE_SSE41
 #endif

+#undef WEBP_MSC_SSE41
+#undef WEBP_MSC_SSE2
+
 // The intrinsics currently cause compiler errors with arm-nacl-gcc and the
 // inline assembly would need to be modified for use with Native Client.
 #if (defined(__ARM_NEON__) || \
@@ -110,8 +112,6 @@ extern "C" {
 #define WEBP_USE_MSA
 #endif

-#endif  /* EMSCRIPTEN */
-
 #ifndef WEBP_DSP_OMIT_C_CODE
 #define WEBP_DSP_OMIT_C_CODE 1
 #endif
@@ -193,6 +193,12 @@ extern "C" {
 #endif
 #endif

+// If 'ptr' is NULL, returns NULL. Otherwise returns 'ptr + off'.
+// Prevents undefined behavior sanitizer nullptr-with-nonzero-offset warning.
+#if !defined(WEBP_OFFSET_PTR)
+#define WEBP_OFFSET_PTR(ptr, off) (((ptr) == NULL) ? NULL : ((ptr) + (off)))
+#endif
+
 // Regularize the definition of WEBP_SWAP_16BIT_CSP (backward compatibility)
 #if !defined(WEBP_SWAP_16BIT_CSP)
 #define WEBP_SWAP_16BIT_CSP 0
@@ -632,6 +638,8 @@ extern void (*WebPPackRGB)(const uint8_t* r, const uint8_t* g, const uint8_t* b,
 extern int (*WebPHasAlpha8b)(const uint8_t* src, int length);
 // This function returns true if src[4*i] contains a value different from 0xff.
 extern int (*WebPHasAlpha32b)(const uint8_t* src, int length);
+// replaces transparent values in src[] by 'color'.
+extern void (*WebPAlphaReplace)(uint32_t* src, int length, uint32_t color);

 // To be called first before using the above.
 void WebPInitAlphaProcessing(void);
@@ -107,62 +107,62 @@ static WEBP_INLINE uint32_t Select(uint32_t a, uint32_t b, uint32_t c) {
 //------------------------------------------------------------------------------
 // Predictors

-static uint32_t Predictor0_C(uint32_t left, const uint32_t* const top) {
+uint32_t VP8LPredictor0_C(uint32_t left, const uint32_t* const top) {
  (void)top;
  (void)left;
  return ARGB_BLACK;
 }
-static uint32_t Predictor1_C(uint32_t left, const uint32_t* const top) {
+uint32_t VP8LPredictor1_C(uint32_t left, const uint32_t* const top) {
  (void)top;
  return left;
 }
-static uint32_t Predictor2_C(uint32_t left, const uint32_t* const top) {
+uint32_t VP8LPredictor2_C(uint32_t left, const uint32_t* const top) {
  (void)left;
  return top[0];
 }
-static uint32_t Predictor3_C(uint32_t left, const uint32_t* const top) {
+uint32_t VP8LPredictor3_C(uint32_t left, const uint32_t* const top) {
  (void)left;
  return top[1];
 }
-static uint32_t Predictor4_C(uint32_t left, const uint32_t* const top) {
+uint32_t VP8LPredictor4_C(uint32_t left, const uint32_t* const top) {
  (void)left;
  return top[-1];
 }
-static uint32_t Predictor5_C(uint32_t left, const uint32_t* const top) {
+uint32_t VP8LPredictor5_C(uint32_t left, const uint32_t* const top) {
  const uint32_t pred = Average3(left, top[0], top[1]);
  return pred;
 }
-static uint32_t Predictor6_C(uint32_t left, const uint32_t* const top) {
+uint32_t VP8LPredictor6_C(uint32_t left, const uint32_t* const top) {
  const uint32_t pred = Average2(left, top[-1]);
  return pred;
 }
-static uint32_t Predictor7_C(uint32_t left, const uint32_t* const top) {
+uint32_t VP8LPredictor7_C(uint32_t left, const uint32_t* const top) {
  const uint32_t pred = Average2(left, top[0]);
  return pred;
 }
-static uint32_t Predictor8_C(uint32_t left, const uint32_t* const top) {
+uint32_t VP8LPredictor8_C(uint32_t left, const uint32_t* const top) {
  const uint32_t pred = Average2(top[-1], top[0]);
  (void)left;
  return pred;
 }
-static uint32_t Predictor9_C(uint32_t left, const uint32_t* const top) {
+uint32_t VP8LPredictor9_C(uint32_t left, const uint32_t* const top) {
  const uint32_t pred = Average2(top[0], top[1]);
  (void)left;
  return pred;
 }
-static uint32_t Predictor10_C(uint32_t left, const uint32_t* const top) {
+uint32_t VP8LPredictor10_C(uint32_t left, const uint32_t* const top) {
  const uint32_t pred = Average4(left, top[-1], top[0], top[1]);
  return pred;
 }
-static uint32_t Predictor11_C(uint32_t left, const uint32_t* const top) {
+uint32_t VP8LPredictor11_C(uint32_t left, const uint32_t* const top) {
  const uint32_t pred = Select(top[0], left, top[-1]);
  return pred;
 }
-static uint32_t Predictor12_C(uint32_t left, const uint32_t* const top) {
+uint32_t VP8LPredictor12_C(uint32_t left, const uint32_t* const top) {
  const uint32_t pred = ClampedAddSubtractFull(left, top[0], top[-1]);
  return pred;
 }
-static uint32_t Predictor13_C(uint32_t left, const uint32_t* const top) {
+uint32_t VP8LPredictor13_C(uint32_t left, const uint32_t* const top) {
  const uint32_t pred = ClampedAddSubtractHalf(left, top[0], top[-1]);
  return pred;
 }
@@ -182,18 +182,18 @@ static void PredictorAdd1_C(const uint32_t* in, const uint32_t* upper,
    out[i] = left = VP8LAddPixels(in[i], left);
  }
 }
-GENERATE_PREDICTOR_ADD(Predictor2_C, PredictorAdd2_C)
-GENERATE_PREDICTOR_ADD(Predictor3_C, PredictorAdd3_C)
-GENERATE_PREDICTOR_ADD(Predictor4_C, PredictorAdd4_C)
-GENERATE_PREDICTOR_ADD(Predictor5_C, PredictorAdd5_C)
-GENERATE_PREDICTOR_ADD(Predictor6_C, PredictorAdd6_C)
-GENERATE_PREDICTOR_ADD(Predictor7_C, PredictorAdd7_C)
-GENERATE_PREDICTOR_ADD(Predictor8_C, PredictorAdd8_C)
-GENERATE_PREDICTOR_ADD(Predictor9_C, PredictorAdd9_C)
-GENERATE_PREDICTOR_ADD(Predictor10_C, PredictorAdd10_C)
-GENERATE_PREDICTOR_ADD(Predictor11_C, PredictorAdd11_C)
-GENERATE_PREDICTOR_ADD(Predictor12_C, PredictorAdd12_C)
-GENERATE_PREDICTOR_ADD(Predictor13_C, PredictorAdd13_C)
+GENERATE_PREDICTOR_ADD(VP8LPredictor2_C, PredictorAdd2_C)
+GENERATE_PREDICTOR_ADD(VP8LPredictor3_C, PredictorAdd3_C)
+GENERATE_PREDICTOR_ADD(VP8LPredictor4_C, PredictorAdd4_C)
+GENERATE_PREDICTOR_ADD(VP8LPredictor5_C, PredictorAdd5_C)
+GENERATE_PREDICTOR_ADD(VP8LPredictor6_C, PredictorAdd6_C)
+GENERATE_PREDICTOR_ADD(VP8LPredictor7_C, PredictorAdd7_C)
+GENERATE_PREDICTOR_ADD(VP8LPredictor8_C, PredictorAdd8_C)
+GENERATE_PREDICTOR_ADD(VP8LPredictor9_C, PredictorAdd9_C)
+GENERATE_PREDICTOR_ADD(VP8LPredictor10_C, PredictorAdd10_C)
+GENERATE_PREDICTOR_ADD(VP8LPredictor11_C, PredictorAdd11_C)
+GENERATE_PREDICTOR_ADD(VP8LPredictor12_C, PredictorAdd12_C)
+GENERATE_PREDICTOR_ADD(VP8LPredictor13_C, PredictorAdd13_C)

 //------------------------------------------------------------------------------

@@ -562,7 +562,6 @@ VP8LPredictorFunc VP8LPredictors[16];

 // exposed plain-C implementations
 VP8LPredictorAddSubFunc VP8LPredictorsAdd_C[16];
-VP8LPredictorFunc VP8LPredictors_C[16];

 VP8LTransformColorInverseFunc VP8LTransformColorInverse;

@@ -600,8 +599,7 @@ extern void VP8LDspInitMSA(void);
 } while (0);

 WEBP_DSP_INIT_FUNC(VP8LDspInit) {
-  COPY_PREDICTOR_ARRAY(Predictor, VP8LPredictors)
-  COPY_PREDICTOR_ARRAY(Predictor, VP8LPredictors_C)
+  COPY_PREDICTOR_ARRAY(VP8LPredictor, VP8LPredictors)
  COPY_PREDICTOR_ARRAY(PredictorAdd, VP8LPredictorsAdd)
  COPY_PREDICTOR_ARRAY(PredictorAdd, VP8LPredictorsAdd_C)

@@ -30,7 +30,22 @@ extern "C" {

 typedef uint32_t (*VP8LPredictorFunc)(uint32_t left, const uint32_t* const top);
 extern VP8LPredictorFunc VP8LPredictors[16];
-extern VP8LPredictorFunc VP8LPredictors_C[16];
+
+uint32_t VP8LPredictor0_C(uint32_t left, const uint32_t* const top);
+uint32_t VP8LPredictor1_C(uint32_t left, const uint32_t* const top);
+uint32_t VP8LPredictor2_C(uint32_t left, const uint32_t* const top);
+uint32_t VP8LPredictor3_C(uint32_t left, const uint32_t* const top);
+uint32_t VP8LPredictor4_C(uint32_t left, const uint32_t* const top);
+uint32_t VP8LPredictor5_C(uint32_t left, const uint32_t* const top);
+uint32_t VP8LPredictor6_C(uint32_t left, const uint32_t* const top);
+uint32_t VP8LPredictor7_C(uint32_t left, const uint32_t* const top);
+uint32_t VP8LPredictor8_C(uint32_t left, const uint32_t* const top);
+uint32_t VP8LPredictor9_C(uint32_t left, const uint32_t* const top);
+uint32_t VP8LPredictor10_C(uint32_t left, const uint32_t* const top);
+uint32_t VP8LPredictor11_C(uint32_t left, const uint32_t* const top);
+uint32_t VP8LPredictor12_C(uint32_t left, const uint32_t* const top);
+uint32_t VP8LPredictor13_C(uint32_t left, const uint32_t* const top);
+
 // These Add/Sub function expects upper[-1] and out[-1] to be readable.
 typedef void (*VP8LPredictorAddSubFunc)(const uint32_t* in,
                                        const uint32_t* upper, int num_pixels,
@@ -184,19 +184,6 @@ static void PREDICTOR_ADD(const uint32_t* in, const uint32_t* upper, \
  }                                                                  \
 }

-// It subtracts the prediction from the input pixel and stores the residual
-// in the output pixel.
-#define GENERATE_PREDICTOR_SUB(PREDICTOR, PREDICTOR_SUB)             \
-static void PREDICTOR_SUB(const uint32_t* in, const uint32_t* upper, \
-                          int num_pixels, uint32_t* out) {           \
-  int x;                                                             \
-  assert(upper != NULL);                                             \
-  for (x = 0; x < num_pixels; ++x) {                                 \
-    const uint32_t pred = (PREDICTOR)(in[x - 1], upper + x);         \
-    out[x] = VP8LSubPixels(in[x], pred);                             \
-  }                                                                  \
-}
-
 #ifdef __cplusplus
 }    // extern "C"
 #endif
--- a/Show More
+++ b/Show More