release: OpenCV 4.2.0

* add cv::compare test when Mat type == CV_16F

* add assertion in cv::compare when src.depth() == CV_16F

* cv::compare assertion minor fix

* core: add more checks

.gitignore

@@ -22,3 +22,4 @@ bin/
 *.log
 *.tlog
 build
+node_modules

3rdparty/carotene/hal/tegra_hal.hpp

@@ -1127,6 +1127,7 @@ inline int TEGRA_FILTERINIT(cvhalFilter2D **context, uchar *kernel_data, size_t
         {
             std::memcpy(ctx->kernel_data + kernel_width * j, kernel_data + kernel_step * j, kernel_width * sizeof(int16_t));
         }
+        break;
     default:
         delete[] ctx->kernel_data;
         delete ctx;
@@ -1243,6 +1244,7 @@ inline int TEGRA_SEPFILTERINIT(cvhalFilter2D **context, int src_type, int dst_ty
         ctx->kernely_data[0]=((int16_t*)kernely_data)[0];
         ctx->kernely_data[1]=((int16_t*)kernely_data)[1];
         ctx->kernely_data[2]=((int16_t*)kernely_data)[2];
+        break;
     default:
         delete ctx;
         return CV_HAL_ERROR_NOT_IMPLEMENTED;

3rdparty/ffmpeg/ffmpeg.cmake

@@ -1,9 +1,9 @@
-# Binaries branch name: ffmpeg/master_20190616
-# Binaries were created for OpenCV: 7b099e0fe2d929e55d6705b6ad510c2c9081606b
-ocv_update(FFMPEG_BINARIES_COMMIT "998718df34e35ea0fa429724875fc3900faa266f")
-ocv_update(FFMPEG_FILE_HASH_BIN32 "f03b47fb809edd2e06b6db135cbd3e49")
-ocv_update(FFMPEG_FILE_HASH_BIN64 "5c4571459570c288d874704244c428b5")
-ocv_update(FFMPEG_FILE_HASH_CMAKE "f710891525a04586d565d0e700e62a9c")
+# Binaries branch name: ffmpeg/master_20191119
+# Binaries were created for OpenCV: 318cba4ce37319ba0b0870aab384d5dc066bb124
+ocv_update(FFMPEG_BINARIES_COMMIT "a66a24e9f410ae05da4baeeb8b451912664ce49c")
+ocv_update(FFMPEG_FILE_HASH_BIN32 "5de6044cad9398549e57bc46fc13908d")
+ocv_update(FFMPEG_FILE_HASH_BIN64 "55c0bc8ad27db00116fabf06508de196")
+ocv_update(FFMPEG_FILE_HASH_CMAKE "ad57c038ba34b868277ccbe6dd0f9602")
 
 function(download_win_ffmpeg script_var)
   set(${script_var} "" PARENT_SCOPE)

3rdparty/libjasper/jasper/jas_config.h

@@ -17,6 +17,12 @@
 #if !defined(JAS_WIN_MSVC_BUILD)
 /* A configure-based build is being used. */
 
+#include <stdio.h>
+
+// uClibc-ng workaround: https://github.com/opencv/opencv/pull/15279
+#ifndef L_tmpnam
+#define L_tmpnam 20
+#endif
 
 
 /* Extra debugging support */

3rdparty/libjpeg/CMakeLists.txt

@@ -15,13 +15,6 @@ else()
   ocv_list_filterout(lib_srcs jmemnobs.c)
 endif()
 
-if(WINRT)
-    add_definitions(-DNO_GETENV)
-    get_directory_property( DirDefs COMPILE_DEFINITIONS )
-    message(STATUS "Adding NO_GETENV to compiler definitions for WINRT:")
-    message(STATUS "   COMPILE_DEFINITIONS = ${DirDefs}")
-endif()
-
 # ----------------------------------------------------------------------------------
 #         Define the library target:
 # ----------------------------------------------------------------------------------

3rdparty/libpng/CMakeLists.txt

@@ -46,8 +46,19 @@ if(";${CPU_BASELINE_FINAL};" MATCHES "SSE2"
   add_definitions(-DPNG_INTEL_SSE)
 endif()
 
+# set definitions and sources for MIPS
+if(";${CPU_BASELINE_FINAL};" MATCHES "MSA")
+    list(APPEND lib_srcs mips/mips_init.c mips/filter_msa_intrinsics.c)
+    add_definitions(-DPNG_MIPS_MSA_OPT=2)
+    ocv_warnings_disable(CMAKE_C_FLAGS -Wshadow)
+else()
+    add_definitions(-DPNG_MIPS_MSA_OPT=0)
+endif()
+
 if(PPC64LE OR PPC64)
-  if(ENABLE_VSX AND NOT PPC64)
+  # VSX3 features are backwards compatible
+  if(";${CPU_BASELINE_FINAL};" MATCHES "VSX.*"
+      AND NOT PPC64)
     list(APPEND lib_srcs powerpc/powerpc_init.c powerpc/filter_vsx_intrinsics.c)
     add_definitions(-DPNG_POWERPC_VSX_OPT=2)
   else()

3rdparty/libpng/mips/filter_msa_intrinsics.c

@@ -0,0 +1,808 @@
+
+/* filter_msa_intrinsics.c - MSA optimised filter functions
+ *
+ * Copyright (c) 2018 Cosmin Truta
+ * Copyright (c) 2016 Glenn Randers-Pehrson
+ * Written by Mandar Sahastrabuddhe, August 2016.
+ *
+ * This code is released under the libpng license.
+ * For conditions of distribution and use, see the disclaimer
+ * and license in png.h
+ */
+
+#include <stdio.h>
+#include <stdint.h>
+#include "../pngpriv.h"
+
+#ifdef PNG_READ_SUPPORTED
+
+/* This code requires -mfpu=msa on the command line: */
+#if PNG_MIPS_MSA_IMPLEMENTATION == 1 /* intrinsics code from pngpriv.h */
+
+#include <msa.h>
+
+/* libpng row pointers are not necessarily aligned to any particular boundary,
+ * however this code will only work with appropriate alignment. mips/mips_init.c
+ * checks for this (and will not compile unless it is done). This code uses
+ * variants of png_aligncast to avoid compiler warnings.
+ */
+#define png_ptr(type,pointer) png_aligncast(type *,pointer)
+#define png_ptrc(type,pointer) png_aligncastconst(const type *,pointer)
+
+/* The following relies on a variable 'temp_pointer' being declared with type
+ * 'type'.  This is written this way just to hide the GCC strict aliasing
+ * warning; note that the code is safe because there never is an alias between
+ * the input and output pointers.
+ */
+#define png_ldr(type,pointer)\
+   (temp_pointer = png_ptr(type,pointer), *temp_pointer)
+
+#if PNG_MIPS_MSA_OPT > 0
+
+#ifdef CLANG_BUILD
+   #define MSA_SRLI_B(a, b)   __msa_srli_b((v16i8) a, b)
+
+   #define LW(psrc)                              \
+   ( {                                           \
+       uint8_t *psrc_lw_m = (uint8_t *) (psrc);  \
+       uint32_t val_m;                           \
+                                                 \
+       asm volatile (                            \
+           "lw  %[val_m],  %[psrc_lw_m]  \n\t"   \
+                                                 \
+           : [val_m] "=r" (val_m)                \
+           : [psrc_lw_m] "m" (*psrc_lw_m)        \
+       );                                        \
+                                                 \
+       val_m;                                    \
+   } )
+
+   #define SH(val, pdst)                         \
+   {                                             \
+       uint8_t *pdst_sh_m = (uint8_t *) (pdst);  \
+       uint16_t val_m = (val);                   \
+                                                 \
+       asm volatile (                            \
+           "sh  %[val_m],  %[pdst_sh_m]  \n\t"   \
+                                                 \
+           : [pdst_sh_m] "=m" (*pdst_sh_m)       \
+           : [val_m] "r" (val_m)                 \
+       );                                        \
+   }
+
+   #define SW(val, pdst)                         \
+   {                                             \
+       uint8_t *pdst_sw_m = (uint8_t *) (pdst);  \
+       uint32_t val_m = (val);                   \
+                                                 \
+       asm volatile (                            \
+           "sw  %[val_m],  %[pdst_sw_m]  \n\t"   \
+                                                 \
+           : [pdst_sw_m] "=m" (*pdst_sw_m)       \
+           : [val_m] "r" (val_m)                 \
+       );                                        \
+   }
+
+       #if (__mips == 64)
+        #define SD(val, pdst)                         \
+        {                                             \
+            uint8_t *pdst_sd_m = (uint8_t *) (pdst);  \
+            uint64_t val_m = (val);                   \
+                                                      \
+            asm volatile (                            \
+                "sd  %[val_m],  %[pdst_sd_m]  \n\t"   \
+                                                      \
+                : [pdst_sd_m] "=m" (*pdst_sd_m)       \
+                : [val_m] "r" (val_m)                 \
+            );                                        \
+        }
+    #else
+        #define SD(val, pdst)                                          \
+        {                                                              \
+            uint8_t *pdst_sd_m = (uint8_t *) (pdst);                   \
+            uint32_t val0_m, val1_m;                                   \
+                                                                       \
+            val0_m = (uint32_t) ((val) & 0x00000000FFFFFFFF);          \
+            val1_m = (uint32_t) (((val) >> 32) & 0x00000000FFFFFFFF);  \
+                                                                       \
+            SW(val0_m, pdst_sd_m);                                     \
+            SW(val1_m, pdst_sd_m + 4);                                 \
+        }
+    #endif
+#else
+   #define MSA_SRLI_B(a, b)   (a >> b)
+
+#if (__mips_isa_rev >= 6)
+   #define LW(psrc)                              \
+   ( {                                           \
+       uint8_t *psrc_lw_m = (uint8_t *) (psrc);  \
+       uint32_t val_m;                           \
+                                                 \
+       asm volatile (                            \
+           "lw  %[val_m],  %[psrc_lw_m]  \n\t"   \
+                                                 \
+           : [val_m] "=r" (val_m)                \
+           : [psrc_lw_m] "m" (*psrc_lw_m)        \
+       );                                        \
+                                                 \
+       val_m;                                    \
+   } )
+
+   #define SH(val, pdst)                         \
+   {                                             \
+       uint8_t *pdst_sh_m = (uint8_t *) (pdst);  \
+       uint16_t val_m = (val);                   \
+                                                 \
+       asm volatile (                            \
+           "sh  %[val_m],  %[pdst_sh_m]  \n\t"   \
+                                                 \
+           : [pdst_sh_m] "=m" (*pdst_sh_m)       \
+           : [val_m] "r" (val_m)                 \
+       );                                        \
+   }
+
+   #define SW(val, pdst)                         \
+   {                                             \
+       uint8_t *pdst_sw_m = (uint8_t *) (pdst);  \
+       uint32_t val_m = (val);                   \
+                                                 \
+       asm volatile (                            \
+           "sw  %[val_m],  %[pdst_sw_m]  \n\t"   \
+                                                 \
+           : [pdst_sw_m] "=m" (*pdst_sw_m)       \
+           : [val_m] "r" (val_m)                 \
+       );                                        \
+   }
+
+   #if (__mips == 64)
+        #define SD(val, pdst)                         \
+        {                                             \
+            uint8_t *pdst_sd_m = (uint8_t *) (pdst);  \
+            uint64_t val_m = (val);                   \
+                                                      \
+            asm volatile (                            \
+                "sd  %[val_m],  %[pdst_sd_m]  \n\t"   \
+                                                      \
+                : [pdst_sd_m] "=m" (*pdst_sd_m)       \
+                : [val_m] "r" (val_m)                 \
+            );                                        \
+        }
+    #else
+        #define SD(val, pdst)                                          \
+        {                                                              \
+            uint8_t *pdst_sd_m = (uint8_t *) (pdst);                   \
+            uint32_t val0_m, val1_m;                                   \
+                                                                       \
+            val0_m = (uint32_t) ((val) & 0x00000000FFFFFFFF);          \
+            val1_m = (uint32_t) (((val) >> 32) & 0x00000000FFFFFFFF);  \
+                                                                       \
+            SW(val0_m, pdst_sd_m);                                     \
+            SW(val1_m, pdst_sd_m + 4);                                 \
+        }
+    #endif
+#else  // !(__mips_isa_rev >= 6)
+   #define LW(psrc)                              \
+   ( {                                           \
+       uint8_t *psrc_lw_m = (uint8_t *) (psrc);  \
+       uint32_t val_m;                           \
+                                                 \
+       asm volatile (                            \
+           "ulw  %[val_m],  %[psrc_lw_m]  \n\t"  \
+                                                 \
+           : [val_m] "=r" (val_m)                \
+           : [psrc_lw_m] "m" (*psrc_lw_m)        \
+       );                                        \
+                                                 \
+       val_m;                                    \
+   } )
+
+   #define SH(val, pdst)                         \
+   {                                             \
+       uint8_t *pdst_sh_m = (uint8_t *) (pdst);  \
+       uint16_t val_m = (val);                   \
+                                                 \
+       asm volatile (                            \
+           "ush  %[val_m],  %[pdst_sh_m]  \n\t"  \
+                                                 \
+           : [pdst_sh_m] "=m" (*pdst_sh_m)       \
+           : [val_m] "r" (val_m)                 \
+       );                                        \
+   }
+
+   #define SW(val, pdst)                         \
+   {                                             \
+       uint8_t *pdst_sw_m = (uint8_t *) (pdst);  \
+       uint32_t val_m = (val);                   \
+                                                 \
+       asm volatile (                            \
+           "usw  %[val_m],  %[pdst_sw_m]  \n\t"  \
+                                                 \
+           : [pdst_sw_m] "=m" (*pdst_sw_m)       \
+           : [val_m] "r" (val_m)                 \
+       );                                        \
+   }
+
+   #define SD(val, pdst)                                          \
+    {                                                              \
+        uint8_t *pdst_sd_m = (uint8_t *) (pdst);                   \
+        uint32_t val0_m, val1_m;                                   \
+                                                                   \
+        val0_m = (uint32_t) ((val) & 0x00000000FFFFFFFF);          \
+        val1_m = (uint32_t) (((val) >> 32) & 0x00000000FFFFFFFF);  \
+                                                                   \
+        SW(val0_m, pdst_sd_m);                                     \
+        SW(val1_m, pdst_sd_m + 4);                                 \
+    }
+
+    #define SW_ZERO(pdst)                      \
+    {                                          \
+        uint8_t *pdst_m = (uint8_t *) (pdst);  \
+                                               \
+        asm volatile (                         \
+            "usw  $0,  %[pdst_m]  \n\t"        \
+                                               \
+            : [pdst_m] "=m" (*pdst_m)          \
+            :                                  \
+        );                                     \
+    }
+#endif  // (__mips_isa_rev >= 6)
+#endif
+
+#define LD_B(RTYPE, psrc) *((RTYPE *) (psrc))
+#define LD_UB(...) LD_B(v16u8, __VA_ARGS__)
+#define LD_B2(RTYPE, psrc, stride, out0, out1)  \
+{                                               \
+    out0 = LD_B(RTYPE, (psrc));                 \
+    out1 = LD_B(RTYPE, (psrc) + stride);        \
+}
+#define LD_UB2(...) LD_B2(v16u8, __VA_ARGS__)
+#define LD_B4(RTYPE, psrc, stride, out0, out1, out2, out3)   \
+{                                                            \
+    LD_B2(RTYPE, (psrc), stride, out0, out1);                \
+    LD_B2(RTYPE, (psrc) + 2 * stride , stride, out2, out3);  \
+}
+#define LD_UB4(...) LD_B4(v16u8, __VA_ARGS__)
+
+#define ST_B(RTYPE, in, pdst) *((RTYPE *) (pdst)) = (in)
+#define ST_UB(...) ST_B(v16u8, __VA_ARGS__)
+#define ST_B2(RTYPE, in0, in1, pdst, stride)  \
+{                                             \
+    ST_B(RTYPE, in0, (pdst));                 \
+    ST_B(RTYPE, in1, (pdst) + stride);        \
+}
+#define ST_UB2(...) ST_B2(v16u8, __VA_ARGS__)
+#define ST_B4(RTYPE, in0, in1, in2, in3, pdst, stride)    \
+{                                                         \
+    ST_B2(RTYPE, in0, in1, (pdst), stride);               \
+    ST_B2(RTYPE, in2, in3, (pdst) + 2 * stride, stride);  \
+}
+#define ST_UB4(...) ST_B4(v16u8, __VA_ARGS__)
+
+#define ADD2(in0, in1, in2, in3, out0, out1)  \
+{                                             \
+    out0 = in0 + in1;                         \
+    out1 = in2 + in3;                         \
+}
+#define ADD3(in0, in1, in2, in3, in4, in5,  \
+             out0, out1, out2)              \
+{                                           \
+    ADD2(in0, in1, in2, in3, out0, out1);   \
+    out2 = in4 + in5;                       \
+}
+#define ADD4(in0, in1, in2, in3, in4, in5, in6, in7,  \
+             out0, out1, out2, out3)                  \
+{                                                     \
+    ADD2(in0, in1, in2, in3, out0, out1);             \
+    ADD2(in4, in5, in6, in7, out2, out3);             \
+}
+
+#define ILVR_B2(RTYPE, in0, in1, in2, in3, out0, out1)      \
+{                                                           \
+    out0 = (RTYPE) __msa_ilvr_b((v16i8) in0, (v16i8) in1);  \
+    out1 = (RTYPE) __msa_ilvr_b((v16i8) in2, (v16i8) in3);  \
+}
+#define ILVR_B2_SH(...) ILVR_B2(v8i16, __VA_ARGS__)
+
+#define HSUB_UB2(RTYPE, in0, in1, out0, out1)                 \
+{                                                             \
+    out0 = (RTYPE) __msa_hsub_u_h((v16u8) in0, (v16u8) in0);  \
+    out1 = (RTYPE) __msa_hsub_u_h((v16u8) in1, (v16u8) in1);  \
+}
+#define HSUB_UB2_SH(...) HSUB_UB2(v8i16, __VA_ARGS__)
+
+#define SLDI_B2_0(RTYPE, in0, in1, out0, out1, slide_val)                 \
+{                                                                         \
+    v16i8 zero_m = { 0 };                                                 \
+    out0 = (RTYPE) __msa_sldi_b((v16i8) zero_m, (v16i8) in0, slide_val);  \
+    out1 = (RTYPE) __msa_sldi_b((v16i8) zero_m, (v16i8) in1, slide_val);  \
+}
+#define SLDI_B2_0_UB(...) SLDI_B2_0(v16u8, __VA_ARGS__)
+
+#define SLDI_B3_0(RTYPE, in0, in1, in2, out0, out1, out2,  slide_val)     \
+{                                                                         \
+    v16i8 zero_m = { 0 };                                                 \
+    SLDI_B2_0(RTYPE, in0, in1, out0, out1, slide_val);                    \
+    out2 = (RTYPE) __msa_sldi_b((v16i8) zero_m, (v16i8) in2, slide_val);  \
+}
+#define SLDI_B3_0_UB(...) SLDI_B3_0(v16u8, __VA_ARGS__)
+
+#define ILVEV_W2(RTYPE, in0, in1, in2, in3, out0, out1)      \
+{                                                            \
+    out0 = (RTYPE) __msa_ilvev_w((v4i32) in1, (v4i32) in0);  \
+    out1 = (RTYPE) __msa_ilvev_w((v4i32) in3, (v4i32) in2);  \
+}
+#define ILVEV_W2_UB(...) ILVEV_W2(v16u8, __VA_ARGS__)
+
+#define ADD_ABS_H3(RTYPE, in0, in1, in2, out0, out1, out2)  \
+{                                                           \
+    RTYPE zero = {0};                                       \
+                                                            \
+    out0 = __msa_add_a_h((v8i16) zero, in0);                \
+    out1 = __msa_add_a_h((v8i16) zero, in1);                \
+    out2 = __msa_add_a_h((v8i16) zero, in2);                \
+}
+#define ADD_ABS_H3_SH(...) ADD_ABS_H3(v8i16, __VA_ARGS__)
+
+#define VSHF_B2(RTYPE, in0, in1, in2, in3, mask0, mask1, out0, out1)       \
+{                                                                          \
+    out0 = (RTYPE) __msa_vshf_b((v16i8) mask0, (v16i8) in1, (v16i8) in0);  \
+    out1 = (RTYPE) __msa_vshf_b((v16i8) mask1, (v16i8) in3, (v16i8) in2);  \
+}
+#define VSHF_B2_UB(...) VSHF_B2(v16u8, __VA_ARGS__)
+
+#define CMP_AND_SELECT(inp0, inp1, inp2, inp3, inp4, inp5, out0)              \
+{                                                                             \
+   v8i16 _sel_h0, _sel_h1;                                                    \
+   v16u8 _sel_b0, _sel_b1;                                                    \
+   _sel_h0 = (v8i16) __msa_clt_u_h((v8u16) inp1, (v8u16) inp0);               \
+   _sel_b0 = (v16u8) __msa_pckev_b((v16i8) _sel_h0, (v16i8) _sel_h0);         \
+   inp0 = (v8i16) __msa_bmnz_v((v16u8) inp0, (v16u8) inp1, (v16u8) _sel_h0);  \
+   inp4 = (v16u8) __msa_bmnz_v(inp3, inp4, _sel_b0);                          \
+   _sel_h1 = (v8i16) __msa_clt_u_h((v8u16) inp2, (v8u16) inp0);               \
+   _sel_b1 = (v16u8) __msa_pckev_b((v16i8) _sel_h1, (v16i8) _sel_h1);         \
+   inp4 = (v16u8) __msa_bmnz_v(inp4, inp5, _sel_b1);                          \
+   out0 += inp4;                                                              \
+}
+
+void png_read_filter_row_up_msa(png_row_infop row_info, png_bytep row,
+                                png_const_bytep prev_row)
+{
+   size_t i, cnt, cnt16, cnt32;
+   size_t istop = row_info->rowbytes;
+   png_bytep rp = row;
+   png_const_bytep pp = prev_row;
+   v16u8 src0, src1, src2, src3, src4, src5, src6, src7;
+
+   for (i = 0; i < (istop >> 6); i++)
+   {
+      LD_UB4(rp, 16, src0, src1, src2, src3);
+      LD_UB4(pp, 16, src4, src5, src6, src7);
+      pp += 64;
+
+	  ADD4(src0, src4, src1, src5, src2, src6, src3, src7,
+	       src0, src1, src2, src3);
+
+      ST_UB4(src0, src1, src2, src3, rp, 16);
+      rp += 64;
+   }
+
+   if (istop & 0x3F)
+   {
+      cnt32 = istop & 0x20;
+      cnt16 = istop & 0x10;
+      cnt = istop & 0xF;
+
+      if(cnt32)
+      {
+         if (cnt16 && cnt)
+         {
+            LD_UB4(rp, 16, src0, src1, src2, src3);
+            LD_UB4(pp, 16, src4, src5, src6, src7);
+
+            ADD4(src0, src4, src1, src5, src2, src6, src3, src7,
+	             src0, src1, src2, src3);
+
+            ST_UB4(src0, src1, src2, src3, rp, 16);
+            rp += 64;
+         }
+         else if (cnt16 || cnt)
+         {
+            LD_UB2(rp, 16, src0, src1);
+            LD_UB2(pp, 16, src4, src5);
+            pp += 32;
+            src2 = LD_UB(rp + 32);
+            src6 = LD_UB(pp);
+
+            ADD3(src0, src4, src1, src5, src2, src6, src0, src1, src2);
+
+            ST_UB2(src0, src1, rp, 16);
+            rp += 32;
+            ST_UB(src2, rp);
+            rp += 16;
+         }
+         else
+         {
+            LD_UB2(rp, 16, src0, src1);
+            LD_UB2(pp, 16, src4, src5);
+
+			ADD2(src0, src4, src1, src5, src0, src1);
+
+            ST_UB2(src0, src1, rp, 16);
+            rp += 32;
+         }
+      }
+      else if (cnt16 && cnt)
+      {
+         LD_UB2(rp, 16, src0, src1);
+         LD_UB2(pp, 16, src4, src5);
+
+         ADD2(src0, src4, src1, src5, src0, src1);
+
+         ST_UB2(src0, src1, rp, 16);
+         rp += 32;
+      }
+      else if (cnt16 || cnt)
+      {
+         src0 = LD_UB(rp);
+         src4 = LD_UB(pp);
+         pp += 16;
+
+         src0 += src4;
+
+         ST_UB(src0, rp);
+         rp += 16;
+      }
+   }
+}
+
+void png_read_filter_row_sub4_msa(png_row_infop row_info, png_bytep row,
+                                  png_const_bytep prev_row)
+{
+   size_t count;
+   size_t istop = row_info->rowbytes;
+   png_bytep src = row;
+   png_bytep nxt = row + 4;
+   int32_t inp0;
+   v16u8 src0, src1, src2, src3, src4;
+   v16u8 dst0, dst1;
+   v16u8 zero = { 0 };
+
+   istop -= 4;
+
+   inp0 = LW(src);
+   src += 4;
+   src0 = (v16u8) __msa_insert_w((v4i32) zero, 0, inp0);
+
+   for (count = 0; count < istop; count += 16)
+   {
+      src1 = LD_UB(src);
+      src += 16;
+
+      src2 = (v16u8) __msa_sldi_b((v16i8) zero, (v16i8) src1, 4);
+      src3 = (v16u8) __msa_sldi_b((v16i8) zero, (v16i8) src1, 8);
+      src4 = (v16u8) __msa_sldi_b((v16i8) zero, (v16i8) src1, 12);
+      src1 += src0;
+      src2 += src1;
+      src3 += src2;
+      src4 += src3;
+      src0 = src4;
+      ILVEV_W2_UB(src1, src2, src3, src4, dst0, dst1);
+      dst0 = (v16u8) __msa_pckev_d((v2i64) dst1, (v2i64) dst0);
+
+      ST_UB(dst0, nxt);
+      nxt += 16;
+   }
+}
+
+void png_read_filter_row_sub3_msa(png_row_infop row_info, png_bytep row,
+                                  png_const_bytep prev_row)
+{
+   size_t count;
+   size_t istop = row_info->rowbytes;
+   png_bytep src = row;
+   png_bytep nxt = row + 3;
+   int64_t out0;
+   int32_t inp0, out1;
+   v16u8 src0, src1, src2, src3, src4, dst0, dst1;
+   v16u8 zero = { 0 };
+   v16i8 mask0 = { 0, 1, 2, 16, 17, 18, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 };
+   v16i8 mask1 = { 0, 1, 2, 3, 4, 5, 16, 17, 18, 19, 20, 21, 0, 0, 0, 0 };
+
+   istop -= 3;
+
+   inp0 = LW(src);
+   src += 3;
+   src0 = (v16u8) __msa_insert_w((v4i32) zero, 0, inp0);
+
+   for (count = 0; count < istop; count += 12)
+   {
+      src1 = LD_UB(src);
+      src += 12;
+
+      src2 = (v16u8) __msa_sldi_b((v16i8) zero, (v16i8) src1, 3);
+      src3 = (v16u8) __msa_sldi_b((v16i8) zero, (v16i8) src1, 6);
+      src4 = (v16u8) __msa_sldi_b((v16i8) zero, (v16i8) src1, 9);
+      src1 += src0;
+      src2 += src1;
+      src3 += src2;
+      src4 += src3;
+      src0 = src4;
+      VSHF_B2_UB(src1, src2, src3, src4, mask0, mask0, dst0, dst1);
+      dst0 = (v16u8) __msa_vshf_b(mask1, (v16i8) dst1, (v16i8) dst0);
+      out0 = __msa_copy_s_d((v2i64) dst0, 0);
+      out1 = __msa_copy_s_w((v4i32) dst0, 2);
+
+      SD(out0, nxt);
+      nxt += 8;
+      SW(out1, nxt);
+      nxt += 4;
+   }
+}
+
+void png_read_filter_row_avg4_msa(png_row_infop row_info, png_bytep row,
+                                  png_const_bytep prev_row)
+{
+   size_t i;
+   png_bytep src = row;
+   png_bytep nxt = row;
+   png_const_bytep pp = prev_row;
+   size_t istop = row_info->rowbytes - 4;
+   int32_t inp0, inp1, out0;
+   v16u8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9, dst0, dst1;
+   v16u8 zero = { 0 };
+
+   inp0 = LW(pp);
+   pp += 4;
+   inp1 = LW(src);
+   src += 4;
+   src0 = (v16u8) __msa_insert_w((v4i32) zero, 0, inp0);
+   src1 = (v16u8) __msa_insert_w((v4i32) zero, 0, inp1);
+   src0 = (v16u8) MSA_SRLI_B(src0, 1);
+   src1 += src0;
+   out0 = __msa_copy_s_w((v4i32) src1, 0);
+   SW(out0, nxt);
+   nxt += 4;
+
+   for (i = 0; i < istop; i += 16)
+   {
+      src2 = LD_UB(pp);
+      pp += 16;
+      src6 = LD_UB(src);
+      src += 16;
+
+      SLDI_B2_0_UB(src2, src6, src3, src7, 4);
+      SLDI_B2_0_UB(src2, src6, src4, src8, 8);
+      SLDI_B2_0_UB(src2, src6, src5, src9, 12);
+      src2 = __msa_ave_u_b(src2, src1);
+      src6 += src2;
+      src3 = __msa_ave_u_b(src3, src6);
+      src7 += src3;
+      src4 = __msa_ave_u_b(src4, src7);
+      src8 += src4;
+      src5 = __msa_ave_u_b(src5, src8);
+      src9 += src5;
+      src1 = src9;
+      ILVEV_W2_UB(src6, src7, src8, src9, dst0, dst1);
+      dst0 = (v16u8) __msa_pckev_d((v2i64) dst1, (v2i64) dst0);
+
+      ST_UB(dst0, nxt);
+      nxt += 16;
+   }
+}
+
+void png_read_filter_row_avg3_msa(png_row_infop row_info, png_bytep row,
+                                  png_const_bytep prev_row)
+{
+   size_t i;
+   png_bytep src = row;
+   png_bytep nxt = row;
+   png_const_bytep pp = prev_row;
+   size_t istop = row_info->rowbytes - 3;
+   int64_t out0;
+   int32_t inp0, inp1, out1;
+   int16_t out2;
+   v16u8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9, dst0, dst1;
+   v16u8 zero = { 0 };
+   v16i8 mask0 = { 0, 1, 2, 16, 17, 18, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 };
+   v16i8 mask1 = { 0, 1, 2, 3, 4, 5, 16, 17, 18, 19, 20, 21, 0, 0, 0, 0 };
+
+   inp0 = LW(pp);
+   pp += 3;
+   inp1 = LW(src);
+   src += 3;
+   src0 = (v16u8) __msa_insert_w((v4i32) zero, 0, inp0);
+   src1 = (v16u8) __msa_insert_w((v4i32) zero, 0, inp1);
+   src0 = (v16u8) MSA_SRLI_B(src0, 1);
+   src1 += src0;
+   out2 = __msa_copy_s_h((v8i16) src1, 0);
+   SH(out2, nxt);
+   nxt += 2;
+   nxt[0] = src1[2];
+   nxt++;
+
+   for (i = 0; i < istop; i += 12)
+   {
+      src2 = LD_UB(pp);
+      pp += 12;
+      src6 = LD_UB(src);
+      src += 12;
+
+      SLDI_B2_0_UB(src2, src6, src3, src7, 3);
+      SLDI_B2_0_UB(src2, src6, src4, src8, 6);
+      SLDI_B2_0_UB(src2, src6, src5, src9, 9);
+      src2 = __msa_ave_u_b(src2, src1);
+      src6 += src2;
+      src3 = __msa_ave_u_b(src3, src6);
+      src7 += src3;
+      src4 = __msa_ave_u_b(src4, src7);
+      src8 += src4;
+      src5 = __msa_ave_u_b(src5, src8);
+      src9 += src5;
+      src1 = src9;
+      VSHF_B2_UB(src6, src7, src8, src9, mask0, mask0, dst0, dst1);
+      dst0 = (v16u8) __msa_vshf_b(mask1, (v16i8) dst1, (v16i8) dst0);
+      out0 = __msa_copy_s_d((v2i64) dst0, 0);
+      out1 = __msa_copy_s_w((v4i32) dst0, 2);
+
+      SD(out0, nxt);
+      nxt += 8;
+      SW(out1, nxt);
+      nxt += 4;
+   }
+}
+
+void png_read_filter_row_paeth4_msa(png_row_infop row_info,
+                                    png_bytep row,
+                                    png_const_bytep prev_row)
+{
+   int32_t count, rp_end;
+   png_bytep nxt;
+   png_const_bytep prev_nxt;
+   int32_t inp0, inp1, res0;
+   v16u8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9;
+   v16u8 src10, src11, src12, src13, dst0, dst1;
+   v8i16 vec0, vec1, vec2;
+   v16u8 zero = { 0 };
+
+   nxt = row;
+   prev_nxt = prev_row;
+
+   inp0 = LW(nxt);
+   inp1 = LW(prev_nxt);
+   prev_nxt += 4;
+   src0 = (v16u8) __msa_insert_w((v4i32) zero, 0, inp0);
+   src1 = (v16u8) __msa_insert_w((v4i32) zero, 0, inp1);
+
+   src1 += src0;
+   res0 = __msa_copy_s_w((v4i32) src1, 0);
+
+   SW(res0, nxt);
+   nxt += 4;
+
+   /* Remainder */
+   rp_end = row_info->rowbytes - 4;
+
+   for (count = 0; count < rp_end; count += 16)
+   {
+      src2 = LD_UB(prev_nxt);
+      prev_nxt += 16;
+      src6 = LD_UB(prev_row);
+      prev_row += 16;
+      src10 = LD_UB(nxt);
+
+      SLDI_B3_0_UB(src2, src6, src10, src3, src7, src11, 4);
+      SLDI_B3_0_UB(src2, src6, src10, src4, src8, src12, 8);
+      SLDI_B3_0_UB(src2, src6, src10, src5, src9, src13, 12);
+      ILVR_B2_SH(src2, src6, src1, src6, vec0, vec1);
+      HSUB_UB2_SH(vec0, vec1, vec0, vec1);
+      vec2 = vec0 + vec1;
+      ADD_ABS_H3_SH(vec0, vec1, vec2, vec0, vec1, vec2);
+      CMP_AND_SELECT(vec0, vec1, vec2, src1, src2, src6, src10);
+      ILVR_B2_SH(src3, src7, src10, src7, vec0, vec1);
+      HSUB_UB2_SH(vec0, vec1, vec0, vec1);
+      vec2 = vec0 + vec1;
+      ADD_ABS_H3_SH(vec0, vec1, vec2, vec0, vec1, vec2);
+      CMP_AND_SELECT(vec0, vec1, vec2, src10, src3, src7, src11);
+      ILVR_B2_SH(src4, src8, src11, src8, vec0, vec1);
+      HSUB_UB2_SH(vec0, vec1, vec0, vec1);
+      vec2 = vec0 + vec1;
+      ADD_ABS_H3_SH(vec0, vec1, vec2, vec0, vec1, vec2);
+      CMP_AND_SELECT(vec0, vec1, vec2, src11, src4, src8, src12);
+      ILVR_B2_SH(src5, src9, src12, src9, vec0, vec1);
+      HSUB_UB2_SH(vec0, vec1, vec0, vec1);
+      vec2 = vec0 + vec1;
+      ADD_ABS_H3_SH(vec0, vec1, vec2, vec0, vec1, vec2);
+      CMP_AND_SELECT(vec0, vec1, vec2, src12, src5, src9, src13);
+      src1 = src13;
+      ILVEV_W2_UB(src10, src11, src12, src1, dst0, dst1);
+      dst0 = (v16u8) __msa_pckev_d((v2i64) dst1, (v2i64) dst0);
+
+      ST_UB(dst0, nxt);
+      nxt += 16;
+   }
+}
+
+void png_read_filter_row_paeth3_msa(png_row_infop row_info,
+                                    png_bytep row,
+                                    png_const_bytep prev_row)
+{
+   int32_t count, rp_end;
+   png_bytep nxt;
+   png_const_bytep prev_nxt;
+   int64_t out0;
+   int32_t inp0, inp1, out1;
+   int16_t out2;
+   v16u8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9, dst0, dst1;
+   v16u8 src10, src11, src12, src13;
+   v8i16 vec0, vec1, vec2;
+   v16u8 zero = { 0 };
+   v16i8 mask0 = { 0, 1, 2, 16, 17, 18, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 };
+   v16i8 mask1 = { 0, 1, 2, 3, 4, 5, 16, 17, 18, 19, 20, 21, 0, 0, 0, 0 };
+
+   nxt = row;
+   prev_nxt = prev_row;
+
+   inp0 = LW(nxt);
+   inp1 = LW(prev_nxt);
+   prev_nxt += 3;
+   src0 = (v16u8) __msa_insert_w((v4i32) zero, 0, inp0);
+   src1 = (v16u8) __msa_insert_w((v4i32) zero, 0, inp1);
+
+   src1 += src0;
+   out2 = __msa_copy_s_h((v8i16) src1, 0);
+
+   SH(out2, nxt);
+   nxt += 2;
+   nxt[0] = src1[2];
+   nxt++;
+
+   /* Remainder */
+   rp_end = row_info->rowbytes - 3;
+
+   for (count = 0; count < rp_end; count += 12)
+   {
+      src2 = LD_UB(prev_nxt);
+      prev_nxt += 12;
+      src6 = LD_UB(prev_row);
+      prev_row += 12;
+      src10 = LD_UB(nxt);
+
+      SLDI_B3_0_UB(src2, src6, src10, src3, src7, src11, 3);
+      SLDI_B3_0_UB(src2, src6, src10, src4, src8, src12, 6);
+      SLDI_B3_0_UB(src2, src6, src10, src5, src9, src13, 9);
+      ILVR_B2_SH(src2, src6, src1, src6, vec0, vec1);
+      HSUB_UB2_SH(vec0, vec1, vec0, vec1);
+      vec2 = vec0 + vec1;
+      ADD_ABS_H3_SH(vec0, vec1, vec2, vec0, vec1, vec2);
+      CMP_AND_SELECT(vec0, vec1, vec2, src1, src2, src6, src10);
+      ILVR_B2_SH(src3, src7, src10, src7, vec0, vec1);
+      HSUB_UB2_SH(vec0, vec1, vec0, vec1);
+      vec2 = vec0 + vec1;
+      ADD_ABS_H3_SH(vec0, vec1, vec2, vec0, vec1, vec2);
+      CMP_AND_SELECT(vec0, vec1, vec2, src10, src3, src7, src11);
+      ILVR_B2_SH(src4, src8, src11, src8, vec0, vec1);
+      HSUB_UB2_SH(vec0, vec1, vec0, vec1);
+      vec2 = vec0 + vec1;
+      ADD_ABS_H3_SH(vec0, vec1, vec2, vec0, vec1, vec2);
+      CMP_AND_SELECT(vec0, vec1, vec2, src11, src4, src8, src12);
+      ILVR_B2_SH(src5, src9, src12, src9, vec0, vec1);
+      HSUB_UB2_SH(vec0, vec1, vec0, vec1);
+      vec2 = vec0 + vec1;
+      ADD_ABS_H3_SH(vec0, vec1, vec2, vec0, vec1, vec2);
+      CMP_AND_SELECT(vec0, vec1, vec2, src12, src5, src9, src13);
+      src1 = src13;
+      VSHF_B2_UB(src10, src11, src12, src13, mask0, mask0, dst0, dst1);
+      dst0 = (v16u8) __msa_vshf_b(mask1, (v16i8) dst1, (v16i8) dst0);
+      out0 = __msa_copy_s_d((v2i64) dst0, 0);
+      out1 = __msa_copy_s_w((v4i32) dst0, 2);
+
+      SD(out0, nxt);
+      nxt += 8;
+      SW(out1, nxt);
+      nxt += 4;
+   }
+}
+
+#endif /* PNG_MIPS_MSA_OPT > 0 */
+#endif /* PNG_MIPS_MSA_IMPLEMENTATION == 1 (intrinsics) */
+#endif /* READ */

3rdparty/libpng/mips/mips_init.c

@@ -0,0 +1,127 @@
+
+/* mips_init.c - MSA optimised filter functions
+ *
+ * Copyright (c) 2018 Cosmin Truta
+ * Copyright (c) 2016 Glenn Randers-Pehrson
+ * Written by Mandar Sahastrabuddhe, 2016.
+ *
+ * This code is released under the libpng license.
+ * For conditions of distribution and use, see the disclaimer
+ * and license in png.h
+ */
+
+/* Below, after checking __linux__, various non-C90 POSIX 1003.1 functions are
+ * called.
+ */
+#define _POSIX_SOURCE 1
+
+#include <stdio.h>
+#include "../pngpriv.h"
+
+#ifdef PNG_READ_SUPPORTED
+
+#if PNG_MIPS_MSA_OPT > 0
+#ifdef PNG_MIPS_MSA_CHECK_SUPPORTED /* Do run-time checks */
+/* WARNING: it is strongly recommended that you do not build libpng with
+ * run-time checks for CPU features if at all possible.  In the case of the MIPS
+ * MSA instructions there is no processor-specific way of detecting the
+ * presence of the required support, therefore run-time detection is extremely
+ * OS specific.
+ *
+ * You may set the macro PNG_MIPS_MSA_FILE to the file name of file containing
+ * a fragment of C source code which defines the png_have_msa function.  There
+ * are a number of implementations in contrib/mips-msa, but the only one that
+ * has partial support is contrib/mips-msa/linux.c - a generic Linux
+ * implementation which reads /proc/cpufino.
+ */
+#ifndef PNG_MIPS_MSA_FILE
+#  ifdef __linux__
+#     define PNG_MIPS_MSA_FILE "contrib/mips-msa/linux.c"
+#  endif
+#endif
+
+#ifdef PNG_MIPS_MSA_FILE
+
+#include <signal.h> /* for sig_atomic_t */
+static int png_have_msa(png_structp png_ptr);
+#include PNG_MIPS_MSA_FILE
+
+#else  /* PNG_MIPS_MSA_FILE */
+#  error "PNG_MIPS_MSA_FILE undefined: no support for run-time MIPS MSA checks"
+#endif /* PNG_MIPS_MSA_FILE */
+#endif /* PNG_MIPS_MSA_CHECK_SUPPORTED */
+
+#ifndef PNG_ALIGNED_MEMORY_SUPPORTED
+#  error "ALIGNED_MEMORY is required; set: -DPNG_ALIGNED_MEMORY_SUPPORTED"
+#endif
+
+void
+png_init_filter_functions_msa(png_structp pp, unsigned int bpp)
+{
+   /* The switch statement is compiled in for MIPS_MSA_API, the call to
+    * png_have_msa is compiled in for MIPS_MSA_CHECK. If both are defined
+    * the check is only performed if the API has not set the MSA option on
+    * or off explicitly. In this case the check controls what happens.
+    */
+
+#ifdef PNG_MIPS_MSA_API_SUPPORTED
+   switch ((pp->options >> PNG_MIPS_MSA) & 3)
+   {
+      case PNG_OPTION_UNSET:
+         /* Allow the run-time check to execute if it has been enabled -
+          * thus both API and CHECK can be turned on.  If it isn't supported
+          * this case will fall through to the 'default' below, which just
+          * returns.
+          */
+#ifdef PNG_MIPS_MSA_CHECK_SUPPORTED
+         {
+            static volatile sig_atomic_t no_msa = -1; /* not checked */
+
+            if (no_msa < 0)
+               no_msa = !png_have_msa(pp);
+
+            if (no_msa)
+               return;
+         }
+#endif /* PNG_MIPS_MSA_CHECK_SUPPORTED */
+         break;
+
+      default: /* OFF or INVALID */
+         return;
+
+      case PNG_OPTION_ON:
+         /* Option turned on */
+         break;
+   }
+   /* IMPORTANT: any new external functions used here must be declared using
+    * PNG_INTERNAL_FUNCTION in ../pngpriv.h.  This is required so that the
+    * 'prefix' option to configure works:
+    *
+    *    ./configure --with-libpng-prefix=foobar_
+    *
+    * Verify you have got this right by running the above command, doing a build
+    * and examining pngprefix.h; it must contain a #define for every external
+    * function you add.  (Notice that this happens automatically for the
+    * initialization function.)
+    */
+   pp->read_filter[PNG_FILTER_VALUE_UP-1] = png_read_filter_row_up_msa;
+
+   if (bpp == 3)
+   {
+      pp->read_filter[PNG_FILTER_VALUE_SUB-1] = png_read_filter_row_sub3_msa;
+      pp->read_filter[PNG_FILTER_VALUE_AVG-1] = png_read_filter_row_avg3_msa;
+      pp->read_filter[PNG_FILTER_VALUE_PAETH-1] = png_read_filter_row_paeth3_msa;
+   }
+   else if (bpp == 4)
+   {
+      pp->read_filter[PNG_FILTER_VALUE_SUB-1] = png_read_filter_row_sub4_msa;
+      pp->read_filter[PNG_FILTER_VALUE_AVG-1] = png_read_filter_row_avg4_msa;
+      pp->read_filter[PNG_FILTER_VALUE_PAETH-1] = png_read_filter_row_paeth4_msa;
+   }
+#else
+   (void)pp;
+   (void)bpp;
+#endif /* PNG_MIPS_MSA_API_SUPPORTED */
+}
+#endif /* PNG_MIPS_MSA_OPT > 0 */
+#endif /* READ */

3rdparty/libpng/patches/20190910-msa-patch.diff

@@ -0,0 +1,53 @@
+diff --git a/3rdparty/libpng/mips/mips_init.c b/3rdparty/libpng/mips/mips_init.c
+index 8dd283deef..6a061cccfa 100644
+--- a/3rdparty/libpng/mips/mips_init.c
++++ b/3rdparty/libpng/mips/mips_init.c
+@@ -73,7 +73,6 @@ png_init_filter_functions_msa(png_structp pp, unsigned int bpp)
+           * this case will fall through to the 'default' below, which just
+           * returns.
+           */
+-#endif /* PNG_MIPS_MSA_API_SUPPORTED */
+ #ifdef PNG_MIPS_MSA_CHECK_SUPPORTED
+          {
+             static volatile sig_atomic_t no_msa = -1; /* not checked */
+@@ -84,12 +83,9 @@ png_init_filter_functions_msa(png_structp pp, unsigned int bpp)
+             if (no_msa)
+                return;
+          }
+-#ifdef PNG_MIPS_MSA_API_SUPPORTED
+-         break;
+-#endif
+ #endif /* PNG_MIPS_MSA_CHECK_SUPPORTED */
++         break;
+ 
+-#ifdef PNG_MIPS_MSA_API_SUPPORTED
+       default: /* OFF or INVALID */
+          return;
+ 
+@@ -97,8 +93,6 @@ png_init_filter_functions_msa(png_structp pp, unsigned int bpp)
+          /* Option turned on */
+          break;
+    }
+-#endif
+-
+    /* IMPORTANT: any new external functions used here must be declared using
+     * PNG_INTERNAL_FUNCTION in ../pngpriv.h.  This is required so that the
+     * 'prefix' option to configure works:
+@@ -118,13 +112,16 @@ png_init_filter_functions_msa(png_structp pp, unsigned int bpp)
+       pp->read_filter[PNG_FILTER_VALUE_AVG-1] = png_read_filter_row_avg3_msa;
+       pp->read_filter[PNG_FILTER_VALUE_PAETH-1] = png_read_filter_row_paeth3_msa;
+    }
+-
+    else if (bpp == 4)
+    {
+       pp->read_filter[PNG_FILTER_VALUE_SUB-1] = png_read_filter_row_sub4_msa;
+       pp->read_filter[PNG_FILTER_VALUE_AVG-1] = png_read_filter_row_avg4_msa;
+       pp->read_filter[PNG_FILTER_VALUE_PAETH-1] = png_read_filter_row_paeth4_msa;
+    }
++#else
++   (void)pp;
++   (void)bpp;
++#endif /* PNG_MIPS_MSA_API_SUPPORTED */
+ }
+ #endif /* PNG_MIPS_MSA_OPT > 0 */
+ #endif /* READ */

3rdparty/libwebp/patches/20190910-msa-asm-patch.diff

@@ -0,0 +1,22 @@
+diff --git a/3rdparty/libwebp/src/dsp/msa_macro.h b/3rdparty/libwebp/src/dsp/msa_macro.h
+index de026a1d9e..a16c0bb300 100644
+--- a/3rdparty/libwebp/src/dsp/msa_macro.h
++++ b/3rdparty/libwebp/src/dsp/msa_macro.h
+@@ -73,7 +73,7 @@
+   static inline TYPE FUNC_NAME(const void* const psrc) {  \
+     const uint8_t* const psrc_m = (const uint8_t*)psrc;   \
+     TYPE val_m;                                           \
+-    asm volatile (                                        \
++    __asm__ volatile (                                        \
+       "" #INSTR " %[val_m], %[psrc_m]  \n\t"              \
+       : [val_m] "=r" (val_m)                              \
+       : [psrc_m] "m" (*psrc_m));                          \
+@@ -86,7 +86,7 @@
+   static inline void FUNC_NAME(TYPE val, void* const pdst) { \
+     uint8_t* const pdst_m = (uint8_t*)pdst;                  \
+     TYPE val_m = val;                                        \
+-    asm volatile (                                           \
++    __asm__ volatile (                                           \
+       " " #INSTR "  %[val_m],  %[pdst_m]  \n\t"              \
+       : [pdst_m] "=m" (*pdst_m)                              \
+       : [val_m] "r" (val_m));                                \

3rdparty/libwebp/src/dec/quant_dec.c

@@ -61,12 +61,17 @@ static const uint16_t kAcTable[128] = {
 
 void VP8ParseQuant(VP8Decoder* const dec) {
   VP8BitReader* const br = &dec->br_;
-  const int base_q0 = VP8GetValue(br, 7);
-  const int dqy1_dc = VP8Get(br) ? VP8GetSignedValue(br, 4) : 0;
-  const int dqy2_dc = VP8Get(br) ? VP8GetSignedValue(br, 4) : 0;
-  const int dqy2_ac = VP8Get(br) ? VP8GetSignedValue(br, 4) : 0;
-  const int dquv_dc = VP8Get(br) ? VP8GetSignedValue(br, 4) : 0;
-  const int dquv_ac = VP8Get(br) ? VP8GetSignedValue(br, 4) : 0;
+  const int base_q0 = VP8GetValue(br, 7, "global-header");
+  const int dqy1_dc = VP8Get(br, "global-header") ?
+       VP8GetSignedValue(br, 4, "global-header") : 0;
+  const int dqy2_dc = VP8Get(br, "global-header") ?
+       VP8GetSignedValue(br, 4, "global-header") : 0;
+  const int dqy2_ac = VP8Get(br, "global-header") ?
+       VP8GetSignedValue(br, 4, "global-header") : 0;
+  const int dquv_dc = VP8Get(br, "global-header") ?
+       VP8GetSignedValue(br, 4, "global-header") : 0;
+  const int dquv_ac = VP8Get(br, "global-header") ?
+       VP8GetSignedValue(br, 4, "global-header") : 0;
 
   const VP8SegmentHeader* const hdr = &dec->segment_hdr_;
   int i;

3rdparty/libwebp/src/dec/tree_dec.c

@@ -296,20 +296,21 @@ static void ParseIntraMode(VP8BitReader* const br,
   // to decode more than 1 keyframe.
   if (dec->segment_hdr_.update_map_) {
     // Hardcoded tree parsing
-    block->segment_ = !VP8GetBit(br, dec->proba_.segments_[0])
-                    ? VP8GetBit(br, dec->proba_.segments_[1])
-                    : 2 + VP8GetBit(br, dec->proba_.segments_[2]);
+    block->segment_ = !VP8GetBit(br, dec->proba_.segments_[0], "segments")
+                    ?  VP8GetBit(br, dec->proba_.segments_[1], "segments")
+                    :  VP8GetBit(br, dec->proba_.segments_[2], "segments") + 2;
   } else {
     block->segment_ = 0;  // default for intra
   }
-  if (dec->use_skip_proba_) block->skip_ = VP8GetBit(br, dec->skip_p_);
+  if (dec->use_skip_proba_) block->skip_ = VP8GetBit(br, dec->skip_p_, "skip");
 
-  block->is_i4x4_ = !VP8GetBit(br, 145);   // decide for B_PRED first
+  block->is_i4x4_ = !VP8GetBit(br, 145, "block-size");
   if (!block->is_i4x4_) {
     // Hardcoded 16x16 intra-mode decision tree.
     const int ymode =
-        VP8GetBit(br, 156) ? (VP8GetBit(br, 128) ? TM_PRED : H_PRED)
-                           : (VP8GetBit(br, 163) ? V_PRED : DC_PRED);
+        VP8GetBit(br, 156, "pred-modes") ?
+            (VP8GetBit(br, 128, "pred-modes") ? TM_PRED : H_PRED) :
+            (VP8GetBit(br, 163, "pred-modes") ? V_PRED : DC_PRED);
     block->imodes_[0] = ymode;
     memset(top, ymode, 4 * sizeof(*top));
     memset(left, ymode, 4 * sizeof(*left));
@@ -323,22 +324,25 @@ static void ParseIntraMode(VP8BitReader* const br,
         const uint8_t* const prob = kBModesProba[top[x]][ymode];
 #if (USE_GENERIC_TREE == 1)
         // Generic tree-parsing
-        int i = kYModesIntra4[VP8GetBit(br, prob[0])];
+        int i = kYModesIntra4[VP8GetBit(br, prob[0], "pred-modes")];
         while (i > 0) {
-          i = kYModesIntra4[2 * i + VP8GetBit(br, prob[i])];
+          i = kYModesIntra4[2 * i + VP8GetBit(br, prob[i], "pred-modes")];
         }
         ymode = -i;
 #else
         // Hardcoded tree parsing
-        ymode = !VP8GetBit(br, prob[0]) ? B_DC_PRED :
-                  !VP8GetBit(br, prob[1]) ? B_TM_PRED :
-                    !VP8GetBit(br, prob[2]) ? B_VE_PRED :
-                      !VP8GetBit(br, prob[3]) ?
-                        (!VP8GetBit(br, prob[4]) ? B_HE_PRED :
-                          (!VP8GetBit(br, prob[5]) ? B_RD_PRED : B_VR_PRED)) :
-                        (!VP8GetBit(br, prob[6]) ? B_LD_PRED :
-                          (!VP8GetBit(br, prob[7]) ? B_VL_PRED :
-                            (!VP8GetBit(br, prob[8]) ? B_HD_PRED : B_HU_PRED)));
+        ymode = !VP8GetBit(br, prob[0], "pred-modes") ? B_DC_PRED :
+                  !VP8GetBit(br, prob[1], "pred-modes") ? B_TM_PRED :
+                    !VP8GetBit(br, prob[2], "pred-modes") ? B_VE_PRED :
+                      !VP8GetBit(br, prob[3], "pred-modes") ?
+                        (!VP8GetBit(br, prob[4], "pred-modes") ? B_HE_PRED :
+                          (!VP8GetBit(br, prob[5], "pred-modes") ? B_RD_PRED
+                                                                 : B_VR_PRED)) :
+                        (!VP8GetBit(br, prob[6], "pred-modes") ? B_LD_PRED :
+                          (!VP8GetBit(br, prob[7], "pred-modes") ? B_VL_PRED :
+                            (!VP8GetBit(br, prob[8], "pred-modes") ? B_HD_PRED
+                                                                   : B_HU_PRED))
+                        );
 #endif  // USE_GENERIC_TREE
         top[x] = ymode;
       }
@@ -348,9 +352,9 @@ static void ParseIntraMode(VP8BitReader* const br,
     }
   }
   // Hardcoded UVMode decision tree
-  block->uvmode_ = !VP8GetBit(br, 142) ? DC_PRED
-                 : !VP8GetBit(br, 114) ? V_PRED
-                 : VP8GetBit(br, 183) ? TM_PRED : H_PRED;
+  block->uvmode_ = !VP8GetBit(br, 142, "pred-modes-uv") ? DC_PRED
+                 : !VP8GetBit(br, 114, "pred-modes-uv") ? V_PRED
+                 : VP8GetBit(br, 183, "pred-modes-uv") ? TM_PRED : H_PRED;
 }
 
 int VP8ParseIntraModeRow(VP8BitReader* const br, VP8Decoder* const dec) {
@@ -514,8 +518,10 @@ void VP8ParseProba(VP8BitReader* const br, VP8Decoder* const dec) {
     for (b = 0; b < NUM_BANDS; ++b) {
       for (c = 0; c < NUM_CTX; ++c) {
         for (p = 0; p < NUM_PROBAS; ++p) {
-          const int v = VP8GetBit(br, CoeffsUpdateProba[t][b][c][p]) ?
-                        VP8GetValue(br, 8) : CoeffsProba0[t][b][c][p];
+          const int v =
+              VP8GetBit(br, CoeffsUpdateProba[t][b][c][p], "global-header") ?
+                        VP8GetValue(br, 8, "global-header") :
+                        CoeffsProba0[t][b][c][p];
           proba->bands_[t][b].probas_[c][p] = v;
         }
       }
@@ -524,9 +530,8 @@ void VP8ParseProba(VP8BitReader* const br, VP8Decoder* const dec) {
       proba->bands_ptr_[t][b] = &proba->bands_[t][kBands[b]];
     }
   }
-  dec->use_skip_proba_ = VP8Get(br);
+  dec->use_skip_proba_ = VP8Get(br, "global-header");
   if (dec->use_skip_proba_) {
-    dec->skip_p_ = VP8GetValue(br, 8);
+    dec->skip_p_ = VP8GetValue(br, 8, "global-header");
   }
 }
-

3rdparty/libwebp/src/dec/vp8_dec.c

@@ -161,23 +161,26 @@ static int ParseSegmentHeader(VP8BitReader* br,
                               VP8SegmentHeader* hdr, VP8Proba* proba) {
   assert(br != NULL);
   assert(hdr != NULL);
-  hdr->use_segment_ = VP8Get(br);
+  hdr->use_segment_ = VP8Get(br, "global-header");
   if (hdr->use_segment_) {
-    hdr->update_map_ = VP8Get(br);
-    if (VP8Get(br)) {   // update data
+    hdr->update_map_ = VP8Get(br, "global-header");
+    if (VP8Get(br, "global-header")) {   // update data
       int s;
-      hdr->absolute_delta_ = VP8Get(br);
+      hdr->absolute_delta_ = VP8Get(br, "global-header");
       for (s = 0; s < NUM_MB_SEGMENTS; ++s) {
-        hdr->quantizer_[s] = VP8Get(br) ? VP8GetSignedValue(br, 7) : 0;
+        hdr->quantizer_[s] = VP8Get(br, "global-header") ?
+            VP8GetSignedValue(br, 7, "global-header") : 0;
       }
       for (s = 0; s < NUM_MB_SEGMENTS; ++s) {
-        hdr->filter_strength_[s] = VP8Get(br) ? VP8GetSignedValue(br, 6) : 0;
+        hdr->filter_strength_[s] = VP8Get(br, "global-header") ?
+            VP8GetSignedValue(br, 6, "global-header") : 0;
       }
     }
     if (hdr->update_map_) {
       int s;
       for (s = 0; s < MB_FEATURE_TREE_PROBS; ++s) {
-        proba->segments_[s] = VP8Get(br) ? VP8GetValue(br, 8) : 255u;
+        proba->segments_[s] = VP8Get(br, "global-header") ?
+            VP8GetValue(br, 8, "global-header") : 255u;
       }
     }
   } else {
@@ -205,7 +208,7 @@ static VP8StatusCode ParsePartitions(VP8Decoder* const dec,
   size_t last_part;
   size_t p;
 
-  dec->num_parts_minus_one_ = (1 << VP8GetValue(br, 2)) - 1;
+  dec->num_parts_minus_one_ = (1 << VP8GetValue(br, 2, "global-header")) - 1;
   last_part = dec->num_parts_minus_one_;
   if (size < 3 * last_part) {
     // we can't even read the sizes with sz[]! That's a failure.
@@ -229,21 +232,21 @@ static VP8StatusCode ParsePartitions(VP8Decoder* const dec,
 // Paragraph 9.4
 static int ParseFilterHeader(VP8BitReader* br, VP8Decoder* const dec) {
   VP8FilterHeader* const hdr = &dec->filter_hdr_;
-  hdr->simple_    = VP8Get(br);
-  hdr->level_     = VP8GetValue(br, 6);
-  hdr->sharpness_ = VP8GetValue(br, 3);
-  hdr->use_lf_delta_ = VP8Get(br);
+  hdr->simple_    = VP8Get(br, "global-header");
+  hdr->level_     = VP8GetValue(br, 6, "global-header");
+  hdr->sharpness_ = VP8GetValue(br, 3, "global-header");
+  hdr->use_lf_delta_ = VP8Get(br, "global-header");
   if (hdr->use_lf_delta_) {
-    if (VP8Get(br)) {   // update lf-delta?
+    if (VP8Get(br, "global-header")) {   // update lf-delta?
       int i;
       for (i = 0; i < NUM_REF_LF_DELTAS; ++i) {
-        if (VP8Get(br)) {
-          hdr->ref_lf_delta_[i] = VP8GetSignedValue(br, 6);
+        if (VP8Get(br, "global-header")) {
+          hdr->ref_lf_delta_[i] = VP8GetSignedValue(br, 6, "global-header");
         }
       }
       for (i = 0; i < NUM_MODE_LF_DELTAS; ++i) {
-        if (VP8Get(br)) {
-          hdr->mode_lf_delta_[i] = VP8GetSignedValue(br, 6);
+        if (VP8Get(br, "global-header")) {
+          hdr->mode_lf_delta_[i] = VP8GetSignedValue(br, 6, "global-header");
         }
       }
     }
@@ -352,8 +355,8 @@ int VP8GetHeaders(VP8Decoder* const dec, VP8Io* const io) {
   buf_size -= frm_hdr->partition_length_;
 
   if (frm_hdr->key_frame_) {
-    pic_hdr->colorspace_ = VP8Get(br);
-    pic_hdr->clamp_type_ = VP8Get(br);
+    pic_hdr->colorspace_ = VP8Get(br, "global-header");
+    pic_hdr->clamp_type_ = VP8Get(br, "global-header");
   }
   if (!ParseSegmentHeader(br, &dec->segment_hdr_, &dec->proba_)) {
     return VP8SetError(dec, VP8_STATUS_BITSTREAM_ERROR,
@@ -378,7 +381,7 @@ int VP8GetHeaders(VP8Decoder* const dec, VP8Io* const io) {
                        "Not a key frame.");
   }
 
-  VP8Get(br);   // ignore the value of update_proba_
+  VP8Get(br, "global-header");   // ignore the value of update_proba_
 
   VP8ParseProba(br, dec);
 
@@ -403,28 +406,28 @@ static const uint8_t kZigzag[16] = {
 // See section 13-2: http://tools.ietf.org/html/rfc6386#section-13.2
 static int GetLargeValue(VP8BitReader* const br, const uint8_t* const p) {
   int v;
-  if (!VP8GetBit(br, p[3])) {
-    if (!VP8GetBit(br, p[4])) {
+  if (!VP8GetBit(br, p[3], "coeffs")) {
+    if (!VP8GetBit(br, p[4], "coeffs")) {
       v = 2;
     } else {
-      v = 3 + VP8GetBit(br, p[5]);
+      v = 3 + VP8GetBit(br, p[5], "coeffs");
     }
   } else {
-    if (!VP8GetBit(br, p[6])) {
-      if (!VP8GetBit(br, p[7])) {
-        v = 5 + VP8GetBit(br, 159);
+    if (!VP8GetBit(br, p[6], "coeffs")) {
+      if (!VP8GetBit(br, p[7], "coeffs")) {
+        v = 5 + VP8GetBit(br, 159, "coeffs");
       } else {
-        v = 7 + 2 * VP8GetBit(br, 165);
-        v += VP8GetBit(br, 145);
+        v = 7 + 2 * VP8GetBit(br, 165, "coeffs");
+        v += VP8GetBit(br, 145, "coeffs");
       }
     } else {
       const uint8_t* tab;
-      const int bit1 = VP8GetBit(br, p[8]);
-      const int bit0 = VP8GetBit(br, p[9 + bit1]);
+      const int bit1 = VP8GetBit(br, p[8], "coeffs");
+      const int bit0 = VP8GetBit(br, p[9 + bit1], "coeffs");
       const int cat = 2 * bit1 + bit0;
       v = 0;
       for (tab = kCat3456[cat]; *tab; ++tab) {
-        v += v + VP8GetBit(br, *tab);
+        v += v + VP8GetBit(br, *tab, "coeffs");
       }
       v += 3 + (8 << cat);
     }
@@ -438,24 +441,24 @@ static int GetCoeffsFast(VP8BitReader* const br,
                          int ctx, const quant_t dq, int n, int16_t* out) {
   const uint8_t* p = prob[n]->probas_[ctx];
   for (; n < 16; ++n) {
-    if (!VP8GetBit(br, p[0])) {
+    if (!VP8GetBit(br, p[0], "coeffs")) {
       return n;  // previous coeff was last non-zero coeff
     }
-    while (!VP8GetBit(br, p[1])) {       // sequence of zero coeffs
+    while (!VP8GetBit(br, p[1], "coeffs")) {       // sequence of zero coeffs
       p = prob[++n]->probas_[0];
       if (n == 16) return 16;
     }
     {        // non zero coeff
       const VP8ProbaArray* const p_ctx = &prob[n + 1]->probas_[0];
       int v;
-      if (!VP8GetBit(br, p[2])) {
+      if (!VP8GetBit(br, p[2], "coeffs")) {
         v = 1;
         p = p_ctx[1];
       } else {
         v = GetLargeValue(br, p);
         p = p_ctx[2];
       }
-      out[kZigzag[n]] = VP8GetSigned(br, v) * dq[n > 0];
+      out[kZigzag[n]] = VP8GetSigned(br, v, "coeffs") * dq[n > 0];
     }
   }
   return 16;
@@ -468,24 +471,24 @@ static int GetCoeffsAlt(VP8BitReader* const br,
                         int ctx, const quant_t dq, int n, int16_t* out) {
   const uint8_t* p = prob[n]->probas_[ctx];
   for (; n < 16; ++n) {
-    if (!VP8GetBitAlt(br, p[0])) {
+    if (!VP8GetBitAlt(br, p[0], "coeffs")) {
       return n;  // previous coeff was last non-zero coeff
     }
-    while (!VP8GetBitAlt(br, p[1])) {       // sequence of zero coeffs
+    while (!VP8GetBitAlt(br, p[1], "coeffs")) {       // sequence of zero coeffs
       p = prob[++n]->probas_[0];
       if (n == 16) return 16;
     }
     {        // non zero coeff
       const VP8ProbaArray* const p_ctx = &prob[n + 1]->probas_[0];
       int v;
-      if (!VP8GetBitAlt(br, p[2])) {
+      if (!VP8GetBitAlt(br, p[2], "coeffs")) {
         v = 1;
         p = p_ctx[1];
       } else {
         v = GetLargeValue(br, p);
         p = p_ctx[2];
       }
-      out[kZigzag[n]] = VP8GetSigned(br, v) * dq[n > 0];
+      out[kZigzag[n]] = VP8GetSigned(br, v, "coeffs") * dq[n > 0];
     }
   }
   return 16;

3rdparty/libwebp/src/dec/vp8i_dec.h

@@ -32,7 +32,7 @@ extern "C" {
 // version numbers
 #define DEC_MAJ_VERSION 1
 #define DEC_MIN_VERSION 0
-#define DEC_REV_VERSION 2
+#define DEC_REV_VERSION 3
 
 // YUV-cache parameters. Cache is 32-bytes wide (= one cacheline).
 // Constraints are: We need to store one 16x16 block of luma samples (y),

3rdparty/libwebp/src/dec/vp8l_dec.c

@@ -362,12 +362,8 @@ static int ReadHuffmanCodes(VP8LDecoder* const dec, int xsize, int ysize,
   VP8LMetadata* const hdr = &dec->hdr_;
   uint32_t* huffman_image = NULL;
   HTreeGroup* htree_groups = NULL;
-  // When reading htrees, some might be unused, as the format allows it.
-  // We will still read them but put them in this htree_group_bogus.
-  HTreeGroup htree_group_bogus;
   HuffmanCode* huffman_tables = NULL;
-  HuffmanCode* huffman_tables_bogus = NULL;
-  HuffmanCode* next = NULL;
+  HuffmanCode* huffman_table = NULL;
   int num_htree_groups = 1;
   int num_htree_groups_max = 1;
   int max_alphabet_size = 0;
@@ -418,12 +414,6 @@ static int ReadHuffmanCodes(VP8LDecoder* const dec, int xsize, int ysize,
         if (*mapped_group == -1) *mapped_group = num_htree_groups++;
         huffman_image[i] = *mapped_group;
       }
-      huffman_tables_bogus = (HuffmanCode*)WebPSafeMalloc(
-          table_size, sizeof(*huffman_tables_bogus));
-      if (huffman_tables_bogus == NULL) {
-        dec->status_ = VP8_STATUS_OUT_OF_MEMORY;
-        goto Error;
-      }
     } else {
       num_htree_groups = num_htree_groups_max;
     }
@@ -453,63 +443,71 @@ static int ReadHuffmanCodes(VP8LDecoder* const dec, int xsize, int ysize,
     goto Error;
   }
 
-  next = huffman_tables;
+  huffman_table = huffman_tables;
   for (i = 0; i < num_htree_groups_max; ++i) {
-    // If the index "i" is unused in the Huffman image, read the coefficients
-    // but store them to a bogus htree_group.
-    const int is_bogus = (mapping != NULL && mapping[i] == -1);
-    HTreeGroup* const htree_group =
-        is_bogus ? &htree_group_bogus :
-        &htree_groups[(mapping == NULL) ? i : mapping[i]];
-    HuffmanCode** const htrees = htree_group->htrees;
-    HuffmanCode* huffman_tables_i = is_bogus ? huffman_tables_bogus : next;
-    int size;
-    int total_size = 0;
-    int is_trivial_literal = 1;
-    int max_bits = 0;
-    for (j = 0; j < HUFFMAN_CODES_PER_META_CODE; ++j) {
-      int alphabet_size = kAlphabetSize[j];
-      htrees[j] = huffman_tables_i;
-      if (j == 0 && color_cache_bits > 0) {
-        alphabet_size += 1 << color_cache_bits;
-      }
-      size =
-          ReadHuffmanCode(alphabet_size, dec, code_lengths, huffman_tables_i);
-      if (size == 0) {
-        goto Error;
-      }
-      if (is_trivial_literal && kLiteralMap[j] == 1) {
-        is_trivial_literal = (huffman_tables_i->bits == 0);
-      }
-      total_size += huffman_tables_i->bits;
-      huffman_tables_i += size;
-      if (j <= ALPHA) {
-        int local_max_bits = code_lengths[0];
-        int k;
-        for (k = 1; k < alphabet_size; ++k) {
-          if (code_lengths[k] > local_max_bits) {
-            local_max_bits = code_lengths[k];
-          }
+    // If the index "i" is unused in the Huffman image, just make sure the
+    // coefficients are valid but do not store them.
+    if (mapping != NULL && mapping[i] == -1) {
+      for (j = 0; j < HUFFMAN_CODES_PER_META_CODE; ++j) {
+        int alphabet_size = kAlphabetSize[j];
+        if (j == 0 && color_cache_bits > 0) {
+          alphabet_size += (1 << color_cache_bits);
+        }
+        // Passing in NULL so that nothing gets filled.
+        if (!ReadHuffmanCode(alphabet_size, dec, code_lengths, NULL)) {
+          goto Error;
         }
-        max_bits += local_max_bits;
       }
-    }
-    if (!is_bogus) next = huffman_tables_i;
-    htree_group->is_trivial_literal = is_trivial_literal;
-    htree_group->is_trivial_code = 0;
-    if (is_trivial_literal) {
-      const int red = htrees[RED][0].value;
-      const int blue = htrees[BLUE][0].value;
-      const int alpha = htrees[ALPHA][0].value;
-      htree_group->literal_arb = ((uint32_t)alpha << 24) | (red << 16) | blue;
-      if (total_size == 0 && htrees[GREEN][0].value < NUM_LITERAL_CODES) {
-        htree_group->is_trivial_code = 1;
-        htree_group->literal_arb |= htrees[GREEN][0].value << 8;
+    } else {
+      HTreeGroup* const htree_group =
+          &htree_groups[(mapping == NULL) ? i : mapping[i]];
+      HuffmanCode** const htrees = htree_group->htrees;
+      int size;
+      int total_size = 0;
+      int is_trivial_literal = 1;
+      int max_bits = 0;
+      for (j = 0; j < HUFFMAN_CODES_PER_META_CODE; ++j) {
+        int alphabet_size = kAlphabetSize[j];
+        htrees[j] = huffman_table;
+        if (j == 0 && color_cache_bits > 0) {
+          alphabet_size += (1 << color_cache_bits);
+        }
+        size = ReadHuffmanCode(alphabet_size, dec, code_lengths, huffman_table);
+        if (size == 0) {
+          goto Error;
+        }
+        if (is_trivial_literal && kLiteralMap[j] == 1) {
+          is_trivial_literal = (huffman_table->bits == 0);
+        }
+        total_size += huffman_table->bits;
+        huffman_table += size;
+        if (j <= ALPHA) {
+          int local_max_bits = code_lengths[0];
+          int k;
+          for (k = 1; k < alphabet_size; ++k) {
+            if (code_lengths[k] > local_max_bits) {
+              local_max_bits = code_lengths[k];
+            }
+          }
+          max_bits += local_max_bits;
+        }
       }
+      htree_group->is_trivial_literal = is_trivial_literal;
+      htree_group->is_trivial_code = 0;
+      if (is_trivial_literal) {
+        const int red = htrees[RED][0].value;
+        const int blue = htrees[BLUE][0].value;
+        const int alpha = htrees[ALPHA][0].value;
+        htree_group->literal_arb = ((uint32_t)alpha << 24) | (red << 16) | blue;
+        if (total_size == 0 && htrees[GREEN][0].value < NUM_LITERAL_CODES) {
+          htree_group->is_trivial_code = 1;
+          htree_group->literal_arb |= htrees[GREEN][0].value << 8;
+        }
+      }
+      htree_group->use_packed_table =
+          !htree_group->is_trivial_code && (max_bits < HUFFMAN_PACKED_BITS);
+      if (htree_group->use_packed_table) BuildPackedTable(htree_group);
     }
-    htree_group->use_packed_table =
-        !htree_group->is_trivial_code && (max_bits < HUFFMAN_PACKED_BITS);
-    if (htree_group->use_packed_table) BuildPackedTable(htree_group);
   }
   ok = 1;
 
@@ -521,7 +519,6 @@ static int ReadHuffmanCodes(VP8LDecoder* const dec, int xsize, int ysize,
 
  Error:
   WebPSafeFree(code_lengths);
-  WebPSafeFree(huffman_tables_bogus);
   WebPSafeFree(mapping);
   if (!ok) {
     WebPSafeFree(huffman_image);

3rdparty/libwebp/src/demux/demux.c

@@ -25,7 +25,7 @@
 
 #define DMUX_MAJ_VERSION 1
 #define DMUX_MIN_VERSION 0
-#define DMUX_REV_VERSION 2
+#define DMUX_REV_VERSION 3
 
 typedef struct {
   size_t start_;        // start location of the data

3rdparty/libwebp/src/dsp/alpha_processing_sse2.c

@@ -214,7 +214,7 @@ static void ApplyAlphaMultiply_SSE2(uint8_t* rgba, int alpha_first,
 // Alpha detection
 
 static int HasAlpha8b_SSE2(const uint8_t* src, int length) {
-  const __m128i all_0xff = _mm_set1_epi8(0xff);
+  const __m128i all_0xff = _mm_set1_epi8((char)0xff);
   int i = 0;
   for (; i + 16 <= length; i += 16) {
     const __m128i v = _mm_loadu_si128((const __m128i*)(src + i));
@@ -228,7 +228,7 @@ static int HasAlpha8b_SSE2(const uint8_t* src, int length) {
 
 static int HasAlpha32b_SSE2(const uint8_t* src, int length) {
   const __m128i alpha_mask = _mm_set1_epi32(0xff);
-  const __m128i all_0xff = _mm_set1_epi8(0xff);
+  const __m128i all_0xff = _mm_set1_epi8((char)0xff);
   int i = 0;
   // We don't know if we can access the last 3 bytes after the last alpha
   // value 'src[4 * length - 4]' (because we don't know if alpha is the first

3rdparty/libwebp/src/dsp/cpu.c

@@ -173,8 +173,8 @@ static int AndroidCPUInfo(CPUFeature feature) {
   const AndroidCpuFamily cpu_family = android_getCpuFamily();
   const uint64_t cpu_features = android_getCpuFeatures();
   if (feature == kNEON) {
-    return (cpu_family == ANDROID_CPU_FAMILY_ARM &&
-            0 != (cpu_features & ANDROID_CPU_ARM_FEATURE_NEON));
+    return cpu_family == ANDROID_CPU_FAMILY_ARM &&
+           (cpu_features & ANDROID_CPU_ARM_FEATURE_NEON) != 0;
   }
   return 0;
 }

3rdparty/libwebp/src/dsp/dec_sse2.c

@@ -326,7 +326,7 @@ static WEBP_INLINE void Update2Pixels_SSE2(__m128i* const pi, __m128i* const qi,
   const __m128i a1_lo = _mm_srai_epi16(*a0_lo, 7);
   const __m128i a1_hi = _mm_srai_epi16(*a0_hi, 7);
   const __m128i delta = _mm_packs_epi16(a1_lo, a1_hi);
-  const __m128i sign_bit = _mm_set1_epi8(0x80);
+  const __m128i sign_bit = _mm_set1_epi8((char)0x80);
   *pi = _mm_adds_epi8(*pi, delta);
   *qi = _mm_subs_epi8(*qi, delta);
   FLIP_SIGN_BIT2(*pi, *qi);
@@ -338,9 +338,9 @@ static WEBP_INLINE void NeedsFilter_SSE2(const __m128i* const p1,
                                          const __m128i* const q0,
                                          const __m128i* const q1,
                                          int thresh, __m128i* const mask) {
-  const __m128i m_thresh = _mm_set1_epi8(thresh);
+  const __m128i m_thresh = _mm_set1_epi8((char)thresh);
   const __m128i t1 = MM_ABS(*p1, *q1);        // abs(p1 - q1)
-  const __m128i kFE = _mm_set1_epi8(0xFE);
+  const __m128i kFE = _mm_set1_epi8((char)0xFE);
   const __m128i t2 = _mm_and_si128(t1, kFE);  // set lsb of each byte to zero
   const __m128i t3 = _mm_srli_epi16(t2, 1);   // abs(p1 - q1) / 2
 
@@ -360,7 +360,7 @@ static WEBP_INLINE void DoFilter2_SSE2(__m128i* const p1, __m128i* const p0,
                                        __m128i* const q0, __m128i* const q1,
                                        int thresh) {
   __m128i a, mask;
-  const __m128i sign_bit = _mm_set1_epi8(0x80);
+  const __m128i sign_bit = _mm_set1_epi8((char)0x80);
   // convert p1/q1 to int8_t (for GetBaseDelta_SSE2)
   const __m128i p1s = _mm_xor_si128(*p1, sign_bit);
   const __m128i q1s = _mm_xor_si128(*q1, sign_bit);
@@ -380,7 +380,7 @@ static WEBP_INLINE void DoFilter4_SSE2(__m128i* const p1, __m128i* const p0,
                                        const __m128i* const mask,
                                        int hev_thresh) {
   const __m128i zero = _mm_setzero_si128();
-  const __m128i sign_bit = _mm_set1_epi8(0x80);
+  const __m128i sign_bit = _mm_set1_epi8((char)0x80);
   const __m128i k64 = _mm_set1_epi8(64);
   const __m128i k3 = _mm_set1_epi8(3);
   const __m128i k4 = _mm_set1_epi8(4);
@@ -427,7 +427,7 @@ static WEBP_INLINE void DoFilter6_SSE2(__m128i* const p2, __m128i* const p1,
                                        const __m128i* const mask,
                                        int hev_thresh) {
   const __m128i zero = _mm_setzero_si128();
-  const __m128i sign_bit = _mm_set1_epi8(0x80);
+  const __m128i sign_bit = _mm_set1_epi8((char)0x80);
   __m128i a, not_hev;
 
   // compute hev mask
@@ -941,7 +941,7 @@ static void VR4_SSE2(uint8_t* dst) {   // Vertical-Right
   const __m128i ABCD0 = _mm_srli_si128(XABCD, 1);
   const __m128i abcd = _mm_avg_epu8(XABCD, ABCD0);
   const __m128i _XABCD = _mm_slli_si128(XABCD, 1);
-  const __m128i IXABCD = _mm_insert_epi16(_XABCD, I | (X << 8), 0);
+  const __m128i IXABCD = _mm_insert_epi16(_XABCD, (short)(I | (X << 8)), 0);
   const __m128i avg1 = _mm_avg_epu8(IXABCD, ABCD0);
   const __m128i lsb = _mm_and_si128(_mm_xor_si128(IXABCD, ABCD0), one);
   const __m128i avg2 = _mm_subs_epu8(avg1, lsb);

3rdparty/libwebp/src/dsp/enc_sse2.c

@@ -777,7 +777,7 @@ static WEBP_INLINE void VR4_SSE2(uint8_t* dst,
   const __m128i ABCD0 = _mm_srli_si128(XABCD, 1);
   const __m128i abcd = _mm_avg_epu8(XABCD, ABCD0);
   const __m128i _XABCD = _mm_slli_si128(XABCD, 1);
-  const __m128i IXABCD = _mm_insert_epi16(_XABCD, I | (X << 8), 0);
+  const __m128i IXABCD = _mm_insert_epi16(_XABCD, (short)(I | (X << 8)), 0);
   const __m128i avg1 = _mm_avg_epu8(IXABCD, ABCD0);
   const __m128i lsb = _mm_and_si128(_mm_xor_si128(IXABCD, ABCD0), one);
   const __m128i avg2 = _mm_subs_epu8(avg1, lsb);

3rdparty/libwebp/src/dsp/filters.c

@@ -33,9 +33,9 @@ static WEBP_INLINE void PredictLine_C(const uint8_t* src, const uint8_t* pred,
                                       uint8_t* dst, int length, int inverse) {
   int i;
   if (inverse) {
-    for (i = 0; i < length; ++i) dst[i] = src[i] + pred[i];
+    for (i = 0; i < length; ++i) dst[i] = (uint8_t)(src[i] + pred[i]);
   } else {
-    for (i = 0; i < length; ++i) dst[i] = src[i] - pred[i];
+    for (i = 0; i < length; ++i) dst[i] = (uint8_t)(src[i] - pred[i]);
   }
 }
 
@@ -155,7 +155,7 @@ static WEBP_INLINE void DoGradientFilter_C(const uint8_t* in,
       const int pred = GradientPredictor_C(preds[w - 1],
                                            preds[w - stride],
                                            preds[w - stride - 1]);
-      out[w] = in[w] + (inverse ? pred : -pred);
+      out[w] = (uint8_t)(in[w] + (inverse ? pred : -pred));
     }
     ++row;
     preds += stride;
@@ -194,7 +194,7 @@ static void HorizontalUnfilter_C(const uint8_t* prev, const uint8_t* in,
   uint8_t pred = (prev == NULL) ? 0 : prev[0];
   int i;
   for (i = 0; i < width; ++i) {
-    out[i] = pred + in[i];
+    out[i] = (uint8_t)(pred + in[i]);
     pred = out[i];
   }
 }
@@ -206,7 +206,7 @@ static void VerticalUnfilter_C(const uint8_t* prev, const uint8_t* in,
     HorizontalUnfilter_C(NULL, in, out, width);
   } else {
     int i;
-    for (i = 0; i < width; ++i) out[i] = prev[i] + in[i];
+    for (i = 0; i < width; ++i) out[i] = (uint8_t)(prev[i] + in[i]);
   }
 }
 #endif  // !WEBP_NEON_OMIT_C_CODE
@@ -220,7 +220,7 @@ static void GradientUnfilter_C(const uint8_t* prev, const uint8_t* in,
     int i;
     for (i = 0; i < width; ++i) {
       top = prev[i];  // need to read this first, in case prev==out
-      left = in[i] + GradientPredictor_C(left, top, top_left);
+      left = (uint8_t)(in[i] + GradientPredictor_C(left, top, top_left));
       top_left = top;
       out[i] = left;
     }

3rdparty/libwebp/src/dsp/filters_sse2.c

@@ -163,7 +163,8 @@ static void GradientPredictDirect_SSE2(const uint8_t* const row,
     _mm_storel_epi64((__m128i*)(out + i), H);
   }
   for (; i < length; ++i) {
-    out[i] = row[i] - GradientPredictor_SSE2(row[i - 1], top[i], top[i - 1]);
+    const int delta = GradientPredictor_SSE2(row[i - 1], top[i], top[i - 1]);
+    out[i] = (uint8_t)(row[i] - delta);
   }
 }
 
@@ -188,7 +189,7 @@ static WEBP_INLINE void DoGradientFilter_SSE2(const uint8_t* in,
 
   // Filter line-by-line.
   while (row < last_row) {
-    out[0] = in[0] - in[-stride];
+    out[0] = (uint8_t)(in[0] - in[-stride]);
     GradientPredictDirect_SSE2(in + 1, in + 1 - stride, out + 1, width - 1);
     ++row;
     in += stride;
@@ -223,7 +224,7 @@ static void HorizontalUnfilter_SSE2(const uint8_t* prev, const uint8_t* in,
                                     uint8_t* out, int width) {
   int i;
   __m128i last;
-  out[0] = in[0] + (prev == NULL ? 0 : prev[0]);
+  out[0] = (uint8_t)(in[0] + (prev == NULL ? 0 : prev[0]));
   if (width <= 1) return;
   last = _mm_set_epi32(0, 0, 0, out[0]);
   for (i = 1; i + 8 <= width; i += 8) {
@@ -238,7 +239,7 @@ static void HorizontalUnfilter_SSE2(const uint8_t* prev, const uint8_t* in,
     _mm_storel_epi64((__m128i*)(out + i), A7);
     last = _mm_srli_epi64(A7, 56);
   }
-  for (; i < width; ++i) out[i] = in[i] + out[i - 1];
+  for (; i < width; ++i) out[i] = (uint8_t)(in[i] + out[i - 1]);
 }
 
 static void VerticalUnfilter_SSE2(const uint8_t* prev, const uint8_t* in,
@@ -259,7 +260,7 @@ static void VerticalUnfilter_SSE2(const uint8_t* prev, const uint8_t* in,
       _mm_storeu_si128((__m128i*)&out[i +  0], C0);
       _mm_storeu_si128((__m128i*)&out[i + 16], C1);
     }
-    for (; i < width; ++i) out[i] = in[i] + prev[i];
+    for (; i < width; ++i) out[i] = (uint8_t)(in[i] + prev[i]);
   }
 }
 
@@ -296,7 +297,8 @@ static void GradientPredictInverse_SSE2(const uint8_t* const in,
       _mm_storel_epi64((__m128i*)&row[i], out);
     }
     for (; i < length; ++i) {
-      row[i] = in[i] + GradientPredictor_SSE2(row[i - 1], top[i], top[i - 1]);
+      const int delta = GradientPredictor_SSE2(row[i - 1], top[i], top[i - 1]);
+      row[i] = (uint8_t)(in[i] + delta);
     }
   }
 }
@@ -306,7 +308,7 @@ static void GradientUnfilter_SSE2(const uint8_t* prev, const uint8_t* in,
   if (prev == NULL) {
     HorizontalUnfilter_SSE2(NULL, in, out, width);
   } else {
-    out[0] = in[0] + prev[0];  // predict from above
+    out[0] = (uint8_t)(in[0] + prev[0]);  // predict from above
     GradientPredictInverse_SSE2(in + 1, prev + 1, out + 1, width - 1);
   }
 }

3rdparty/libwebp/src/dsp/lossless.c

@@ -270,14 +270,14 @@ void VP8LTransformColorInverse_C(const VP8LMultipliers* const m,
   int i;
   for (i = 0; i < num_pixels; ++i) {
     const uint32_t argb = src[i];
-    const uint32_t green = argb >> 8;
+    const int8_t green = (int8_t)(argb >> 8);
     const uint32_t red = argb >> 16;
     int new_red = red & 0xff;
     int new_blue = argb & 0xff;
     new_red += ColorTransformDelta(m->green_to_red_, green);
     new_red &= 0xff;
     new_blue += ColorTransformDelta(m->green_to_blue_, green);
-    new_blue += ColorTransformDelta(m->red_to_blue_, new_red);
+    new_blue += ColorTransformDelta(m->red_to_blue_, (int8_t)new_red);
     new_blue &= 0xff;
     dst[i] = (argb & 0xff00ff00u) | (new_red << 16) | (new_blue);
   }

3rdparty/libwebp/src/dsp/lossless_enc.c

@@ -515,13 +515,17 @@ static WEBP_INLINE int ColorTransformDelta(int8_t color_pred, int8_t color) {
   return ((int)color_pred * color) >> 5;
 }
 
+static WEBP_INLINE int8_t U32ToS8(uint32_t v) {
+  return (int8_t)(v & 0xff);
+}
+
 void VP8LTransformColor_C(const VP8LMultipliers* const m, uint32_t* data,
                           int num_pixels) {
   int i;
   for (i = 0; i < num_pixels; ++i) {
     const uint32_t argb = data[i];
-    const uint32_t green = argb >> 8;
-    const uint32_t red = argb >> 16;
+    const int8_t green = U32ToS8(argb >>  8);
+    const int8_t red   = U32ToS8(argb >> 16);
     int new_red = red & 0xff;
     int new_blue = argb & 0xff;
     new_red -= ColorTransformDelta(m->green_to_red_, green);
@@ -535,7 +539,7 @@ void VP8LTransformColor_C(const VP8LMultipliers* const m, uint32_t* data,
 
 static WEBP_INLINE uint8_t TransformColorRed(uint8_t green_to_red,
                                              uint32_t argb) {
-  const uint32_t green = argb >> 8;
+  const int8_t green = U32ToS8(argb >> 8);
   int new_red = argb >> 16;
   new_red -= ColorTransformDelta(green_to_red, green);
   return (new_red & 0xff);
@@ -544,9 +548,9 @@ static WEBP_INLINE uint8_t TransformColorRed(uint8_t green_to_red,
 static WEBP_INLINE uint8_t TransformColorBlue(uint8_t green_to_blue,
                                               uint8_t red_to_blue,
                                               uint32_t argb) {
-  const uint32_t green = argb >> 8;
-  const uint32_t red = argb >> 16;
-  uint8_t new_blue = argb;
+  const int8_t green = U32ToS8(argb >>  8);
+  const int8_t red   = U32ToS8(argb >> 16);
+  uint8_t new_blue = argb & 0xff;
   new_blue -= ColorTransformDelta(green_to_blue, green);
   new_blue -= ColorTransformDelta(red_to_blue, red);
   return (new_blue & 0xff);
@@ -558,7 +562,7 @@ void VP8LCollectColorRedTransforms_C(const uint32_t* argb, int stride,
   while (tile_height-- > 0) {
     int x;
     for (x = 0; x < tile_width; ++x) {
-      ++histo[TransformColorRed(green_to_red, argb[x])];
+      ++histo[TransformColorRed((uint8_t)green_to_red, argb[x])];
     }
     argb += stride;
   }
@@ -571,7 +575,8 @@ void VP8LCollectColorBlueTransforms_C(const uint32_t* argb, int stride,
   while (tile_height-- > 0) {
     int x;
     for (x = 0; x < tile_width; ++x) {
-      ++histo[TransformColorBlue(green_to_blue, red_to_blue, argb[x])];
+      ++histo[TransformColorBlue((uint8_t)green_to_blue, (uint8_t)red_to_blue,
+                                 argb[x])];
     }
     argb += stride;
   }

3rdparty/libwebp/src/dsp/lossless_enc_sse2.c

@@ -363,7 +363,7 @@ static void BundleColorMap_SSE2(const uint8_t* const row, int width, int xbits,
   assert(xbits <= 3);
   switch (xbits) {
     case 0: {
-      const __m128i ff = _mm_set1_epi16(0xff00);
+      const __m128i ff = _mm_set1_epi16((short)0xff00);
       const __m128i zero = _mm_setzero_si128();
       // Store 0xff000000 | (row[x] << 8).
       for (x = 0; x + 16 <= width; x += 16, dst += 16) {
@@ -382,7 +382,7 @@ static void BundleColorMap_SSE2(const uint8_t* const row, int width, int xbits,
       break;
     }
     case 1: {
-      const __m128i ff = _mm_set1_epi16(0xff00);
+      const __m128i ff = _mm_set1_epi16((short)0xff00);
       const __m128i mul = _mm_set1_epi16(0x110);
       for (x = 0; x + 16 <= width; x += 16, dst += 8) {
         // 0a0b | (where a/b are 4 bits).

3rdparty/libwebp/src/dsp/lossless_enc_sse41.c

@@ -51,9 +51,9 @@ static void CollectColorBlueTransforms_SSE41(const uint32_t* argb, int stride,
                                              int histo[]) {
   const __m128i mults_r = _mm_set1_epi16(CST_5b(red_to_blue));
   const __m128i mults_g = _mm_set1_epi16(CST_5b(green_to_blue));
-  const __m128i mask_g = _mm_set1_epi16(0xff00);   // green mask
-  const __m128i mask_gb = _mm_set1_epi32(0xffff);  // green/blue mask
-  const __m128i mask_b = _mm_set1_epi16(0x00ff);   // blue mask
+  const __m128i mask_g = _mm_set1_epi16((short)0xff00);   // green mask
+  const __m128i mask_gb = _mm_set1_epi32(0xffff);         // green/blue mask
+  const __m128i mask_b = _mm_set1_epi16(0x00ff);          // blue mask
   const __m128i shuffler_lo = _mm_setr_epi8(-1, 2, -1, 6, -1, 10, -1, 14, -1,
                                             -1, -1, -1, -1, -1, -1, -1);
   const __m128i shuffler_hi = _mm_setr_epi8(-1, -1, -1, -1, -1, -1, -1, -1, -1,

3rdparty/libwebp/src/dsp/msa_macro.h

@@ -73,7 +73,7 @@
   static inline TYPE FUNC_NAME(const void* const psrc) {  \
     const uint8_t* const psrc_m = (const uint8_t*)psrc;   \
     TYPE val_m;                                           \
-    asm volatile (                                        \
+    __asm__ volatile (                                        \
       "" #INSTR " %[val_m], %[psrc_m]  \n\t"              \
       : [val_m] "=r" (val_m)                              \
       : [psrc_m] "m" (*psrc_m));                          \
@@ -86,7 +86,7 @@
   static inline void FUNC_NAME(TYPE val, void* const pdst) { \
     uint8_t* const pdst_m = (uint8_t*)pdst;                  \
     TYPE val_m = val;                                        \
-    asm volatile (                                           \
+    __asm__ volatile (                                           \
       " " #INSTR "  %[val_m],  %[pdst_m]  \n\t"              \
       : [pdst_m] "=m" (*pdst_m)                              \
       : [val_m] "r" (val_m));                                \

3rdparty/libwebp/src/dsp/quant.h

@@ -10,6 +10,8 @@
 #ifndef WEBP_DSP_QUANT_H_
 #define WEBP_DSP_QUANT_H_
 
+#include <string.h>
+
 #include "src/dsp/dsp.h"
 #include "src/webp/types.h"
 
@@ -67,4 +69,17 @@ static WEBP_INLINE int IsFlat(const int16_t* levels, int num_blocks,
 #endif  // defined(WEBP_USE_NEON) && !defined(WEBP_ANDROID_NEON) &&
         // !defined(WEBP_HAVE_NEON_RTCD)
 
+static WEBP_INLINE int IsFlatSource16(const uint8_t* src) {
+  const uint32_t v = src[0] * 0x01010101u;
+  int i;
+  for (i = 0; i < 16; ++i) {
+    if (memcmp(src + 0, &v, 4) || memcmp(src +  4, &v, 4) ||
+        memcmp(src + 8, &v, 4) || memcmp(src + 12, &v, 4)) {
+      return 0;
+    }
+    src += BPS;
+  }
+  return 1;
+}
+
 #endif  // WEBP_DSP_QUANT_H_

3rdparty/libwebp/src/dsp/rescaler.c

@@ -109,8 +109,7 @@ void WebPRescalerExportRowExpand_C(WebPRescaler* const wrk) {
     for (x_out = 0; x_out < x_out_max; ++x_out) {
       const uint32_t J = frow[x_out];
       const int v = (int)MULT_FIX(J, wrk->fy_scale);
-      assert(v >= 0 && v <= 255);
-      dst[x_out] = v;
+      dst[x_out] = (v > 255) ? 255u : (uint8_t)v;
     }
   } else {
     const uint32_t B = WEBP_RESCALER_FRAC(-wrk->y_accum, wrk->y_sub);
@@ -120,8 +119,7 @@ void WebPRescalerExportRowExpand_C(WebPRescaler* const wrk) {
                        + (uint64_t)B * irow[x_out];
       const uint32_t J = (uint32_t)((I + ROUNDER) >> WEBP_RESCALER_RFIX);
       const int v = (int)MULT_FIX(J, wrk->fy_scale);
-      assert(v >= 0 && v <= 255);
-      dst[x_out] = v;
+      dst[x_out] = (v > 255) ? 255u : (uint8_t)v;
     }
   }
 }
@@ -138,17 +136,15 @@ void WebPRescalerExportRowShrink_C(WebPRescaler* const wrk) {
   assert(!wrk->y_expand);
   if (yscale) {
     for (x_out = 0; x_out < x_out_max; ++x_out) {
-      const uint32_t frac = (uint32_t)MULT_FIX(frow[x_out], yscale);
-      const int v = (int)MULT_FIX_FLOOR(irow[x_out] - frac, wrk->fxy_scale);
-      assert(v >= 0 && v <= 255);
-      dst[x_out] = v;
+      const uint32_t frac = (uint32_t)MULT_FIX_FLOOR(frow[x_out], yscale);
+      const int v = (int)MULT_FIX(irow[x_out] - frac, wrk->fxy_scale);
+      dst[x_out] = (v > 255) ? 255u : (uint8_t)v;
       irow[x_out] = frac;   // new fractional start
     }
   } else {
     for (x_out = 0; x_out < x_out_max; ++x_out) {
       const int v = (int)MULT_FIX(irow[x_out], wrk->fxy_scale);
-      assert(v >= 0 && v <= 255);
-      dst[x_out] = v;
+      dst[x_out] = (v > 255) ? 255u : (uint8_t)v;
       irow[x_out] = 0;
     }
   }

3rdparty/libwebp/src/dsp/rescaler_mips_dsp_r2.c

@@ -107,10 +107,9 @@ static void ExportRowShrink_MIPSdspR2(WebPRescaler* const wrk) {
       );
     }
     for (i = 0; i < (x_out_max & 0x3); ++i) {
-      const uint32_t frac = (uint32_t)MULT_FIX(*frow++, yscale);
-      const int v = (int)MULT_FIX_FLOOR(*irow - frac, wrk->fxy_scale);
-      assert(v >= 0 && v <= 255);
-      *dst++ = v;
+      const uint32_t frac = (uint32_t)MULT_FIX_FLOOR(*frow++, yscale);
+      const int v = (int)MULT_FIX(*irow - frac, wrk->fxy_scale);
+      *dst++ = (v > 255) ? 255u : (uint8_t)v;
       *irow++ = frac;   // new fractional start
     }
   } else {
@@ -157,8 +156,7 @@ static void ExportRowShrink_MIPSdspR2(WebPRescaler* const wrk) {
     }
     for (i = 0; i < (x_out_max & 0x3); ++i) {
       const int v = (int)MULT_FIX_FLOOR(*irow, wrk->fxy_scale);
-      assert(v >= 0 && v <= 255);
-      *dst++ = v;
+      *dst++ = (v > 255) ? 255u : (uint8_t)v;
       *irow++ = 0;
     }
   }
@@ -219,8 +217,7 @@ static void ExportRowExpand_MIPSdspR2(WebPRescaler* const wrk) {
     for (i = 0; i < (x_out_max & 0x3); ++i) {
       const uint32_t J = *frow++;
       const int v = (int)MULT_FIX(J, wrk->fy_scale);
-      assert(v >= 0 && v <= 255);
-      *dst++ = v;
+      *dst++ = (v > 255) ? 255u : (uint8_t)v;
     }
   } else {
     const uint32_t B = WEBP_RESCALER_FRAC(-wrk->y_accum, wrk->y_sub);
@@ -291,8 +288,7 @@ static void ExportRowExpand_MIPSdspR2(WebPRescaler* const wrk) {
                        + (uint64_t)B * *irow++;
       const uint32_t J = (uint32_t)((I + ROUNDER) >> WEBP_RESCALER_RFIX);
       const int v = (int)MULT_FIX(J, wrk->fy_scale);
-      assert(v >= 0 && v <= 255);
-      *dst++ = v;
+      *dst++ = (v > 255) ? 255u : (uint8_t)v;
     }
   }
 }

3rdparty/libwebp/src/dsp/rescaler_msa.c

@@ -166,8 +166,7 @@ static WEBP_INLINE void ExportRowExpand_0(const uint32_t* frow, uint8_t* dst,
     for (x_out = 0; x_out < length; ++x_out) {
       const uint32_t J = frow[x_out];
       const int v = (int)MULT_FIX(J, wrk->fy_scale);
-      assert(v >= 0 && v <= 255);
-      dst[x_out] = v;
+      dst[x_out] = (v > 255) ? 255u : (uint8_t)v;
     }
   }
 }
@@ -241,8 +240,7 @@ static WEBP_INLINE void ExportRowExpand_1(const uint32_t* frow, uint32_t* irow,
                        + (uint64_t)B * irow[x_out];
       const uint32_t J = (uint32_t)((I + ROUNDER) >> WEBP_RESCALER_RFIX);
       const int v = (int)MULT_FIX(J, wrk->fy_scale);
-      assert(v >= 0 && v <= 255);
-      dst[x_out] = v;
+      dst[x_out] = (v > 255) ? 255u : (uint8_t)v;
     }
   }
 }
@@ -342,10 +340,9 @@ static WEBP_INLINE void ExportRowShrink_0(const uint32_t* frow, uint32_t* irow,
       length -= 4;
     }
     for (x_out = 0; x_out < length; ++x_out) {
-      const uint32_t frac = (uint32_t)MULT_FIX(frow[x_out], yscale);
-      const int v = (int)MULT_FIX_FLOOR(irow[x_out] - frac, wrk->fxy_scale);
-      assert(v >= 0 && v <= 255);
-      dst[x_out] = v;
+      const uint32_t frac = (uint32_t)MULT_FIX_FLOOR(frow[x_out], yscale);
+      const int v = (int)MULT_FIX(irow[x_out] - frac, wrk->fxy_scale);
+      dst[x_out] = (v > 255) ? 255u : (uint8_t)v;
       irow[x_out] = frac;
     }
   }
@@ -406,8 +403,7 @@ static WEBP_INLINE void ExportRowShrink_1(uint32_t* irow, uint8_t* dst,
     }
     for (x_out = 0; x_out < length; ++x_out) {
       const int v = (int)MULT_FIX(irow[x_out], wrk->fxy_scale);
-      assert(v >= 0 && v <= 255);
-      dst[x_out] = v;
+      dst[x_out] = (v > 255) ? 255u : (uint8_t)v;
       irow[x_out] = 0;
     }
   }

3rdparty/libwebp/src/dsp/rescaler_neon.c

@@ -81,14 +81,13 @@ static void RescalerExportRowExpand_NEON(WebPRescaler* const wrk) {
       const uint32x4_t B1 = MULT_FIX(A1, fy_scale_half);
       const uint16x4_t C0 = vmovn_u32(B0);
       const uint16x4_t C1 = vmovn_u32(B1);
-      const uint8x8_t D = vmovn_u16(vcombine_u16(C0, C1));
+      const uint8x8_t D = vqmovn_u16(vcombine_u16(C0, C1));
       vst1_u8(dst + x_out, D);
     }
     for (; x_out < x_out_max; ++x_out) {
       const uint32_t J = frow[x_out];
       const int v = (int)MULT_FIX_C(J, fy_scale);
-      assert(v >= 0 && v <= 255);
-      dst[x_out] = v;
+      dst[x_out] = (v > 255) ? 255u : (uint8_t)v;
     }
   } else {
     const uint32_t B = WEBP_RESCALER_FRAC(-wrk->y_accum, wrk->y_sub);
@@ -102,7 +101,7 @@ static void RescalerExportRowExpand_NEON(WebPRescaler* const wrk) {
       const uint32x4_t D1 = MULT_FIX(C1, fy_scale_half);
       const uint16x4_t E0 = vmovn_u32(D0);
       const uint16x4_t E1 = vmovn_u32(D1);
-      const uint8x8_t F = vmovn_u16(vcombine_u16(E0, E1));
+      const uint8x8_t F = vqmovn_u16(vcombine_u16(E0, E1));
       vst1_u8(dst + x_out, F);
     }
     for (; x_out < x_out_max; ++x_out) {
@@ -110,8 +109,7 @@ static void RescalerExportRowExpand_NEON(WebPRescaler* const wrk) {
                        + (uint64_t)B * irow[x_out];
       const uint32_t J = (uint32_t)((I + ROUNDER) >> WEBP_RESCALER_RFIX);
       const int v = (int)MULT_FIX_C(J, fy_scale);
-      assert(v >= 0 && v <= 255);
-      dst[x_out] = v;
+      dst[x_out] = (v > 255) ? 255u : (uint8_t)v;
     }
   }
 }
@@ -135,23 +133,22 @@ static void RescalerExportRowShrink_NEON(WebPRescaler* const wrk) {
     for (x_out = 0; x_out < max_span; x_out += 8) {
       LOAD_32x8(frow + x_out, in0, in1);
       LOAD_32x8(irow + x_out, in2, in3);
-      const uint32x4_t A0 = MULT_FIX(in0, yscale_half);
-      const uint32x4_t A1 = MULT_FIX(in1, yscale_half);
+      const uint32x4_t A0 = MULT_FIX_FLOOR(in0, yscale_half);
+      const uint32x4_t A1 = MULT_FIX_FLOOR(in1, yscale_half);
       const uint32x4_t B0 = vqsubq_u32(in2, A0);
       const uint32x4_t B1 = vqsubq_u32(in3, A1);
-      const uint32x4_t C0 = MULT_FIX_FLOOR(B0, fxy_scale_half);
-      const uint32x4_t C1 = MULT_FIX_FLOOR(B1, fxy_scale_half);
+      const uint32x4_t C0 = MULT_FIX(B0, fxy_scale_half);
+      const uint32x4_t C1 = MULT_FIX(B1, fxy_scale_half);
       const uint16x4_t D0 = vmovn_u32(C0);
       const uint16x4_t D1 = vmovn_u32(C1);
-      const uint8x8_t E = vmovn_u16(vcombine_u16(D0, D1));
+      const uint8x8_t E = vqmovn_u16(vcombine_u16(D0, D1));
       vst1_u8(dst + x_out, E);
       STORE_32x8(A0, A1, irow + x_out);
     }
     for (; x_out < x_out_max; ++x_out) {
-      const uint32_t frac = (uint32_t)MULT_FIX_C(frow[x_out], yscale);
-      const int v = (int)MULT_FIX_FLOOR_C(irow[x_out] - frac, fxy_scale);
-      assert(v >= 0 && v <= 255);
-      dst[x_out] = v;
+      const uint32_t frac = (uint32_t)MULT_FIX_FLOOR_C(frow[x_out], yscale);
+      const int v = (int)MULT_FIX_C(irow[x_out] - frac, fxy_scale);
+      dst[x_out] = (v > 255) ? 255u : (uint8_t)v;
       irow[x_out] = frac;   // new fractional start
     }
   } else {
@@ -161,14 +158,13 @@ static void RescalerExportRowShrink_NEON(WebPRescaler* const wrk) {
       const uint32x4_t A1 = MULT_FIX(in1, fxy_scale_half);
       const uint16x4_t B0 = vmovn_u32(A0);
       const uint16x4_t B1 = vmovn_u32(A1);
-      const uint8x8_t C = vmovn_u16(vcombine_u16(B0, B1));
+      const uint8x8_t C = vqmovn_u16(vcombine_u16(B0, B1));
       vst1_u8(dst + x_out, C);
       STORE_32x8(zero, zero, irow + x_out);
     }
     for (; x_out < x_out_max; ++x_out) {
       const int v = (int)MULT_FIX_C(irow[x_out], fxy_scale);
-      assert(v >= 0 && v <= 255);
-      dst[x_out] = v;
+      dst[x_out] = (v > 255) ? 255u : (uint8_t)v;
       irow[x_out] = 0;
     }
   }

3rdparty/libwebp/src/dsp/rescaler_sse2.c

@@ -225,35 +225,6 @@ static WEBP_INLINE void ProcessRow_SSE2(const __m128i* const A0,
   _mm_storel_epi64((__m128i*)dst, G);
 }
 
-static WEBP_INLINE void ProcessRow_Floor_SSE2(const __m128i* const A0,
-                                              const __m128i* const A1,
-                                              const __m128i* const A2,
-                                              const __m128i* const A3,
-                                              const __m128i* const mult,
-                                              uint8_t* const dst) {
-  const __m128i mask = _mm_set_epi32(0xffffffffu, 0, 0xffffffffu, 0);
-  const __m128i B0 = _mm_mul_epu32(*A0, *mult);
-  const __m128i B1 = _mm_mul_epu32(*A1, *mult);
-  const __m128i B2 = _mm_mul_epu32(*A2, *mult);
-  const __m128i B3 = _mm_mul_epu32(*A3, *mult);
-  const __m128i D0 = _mm_srli_epi64(B0, WEBP_RESCALER_RFIX);
-  const __m128i D1 = _mm_srli_epi64(B1, WEBP_RESCALER_RFIX);
-#if (WEBP_RESCALER_RFIX < 32)
-  const __m128i D2 =
-      _mm_and_si128(_mm_slli_epi64(B2, 32 - WEBP_RESCALER_RFIX), mask);
-  const __m128i D3 =
-      _mm_and_si128(_mm_slli_epi64(B3, 32 - WEBP_RESCALER_RFIX), mask);
-#else
-  const __m128i D2 = _mm_and_si128(B2, mask);
-  const __m128i D3 = _mm_and_si128(B3, mask);
-#endif
-  const __m128i E0 = _mm_or_si128(D0, D2);
-  const __m128i E1 = _mm_or_si128(D1, D3);
-  const __m128i F = _mm_packs_epi32(E0, E1);
-  const __m128i G = _mm_packus_epi16(F, F);
-  _mm_storel_epi64((__m128i*)dst, G);
-}
-
 static void RescalerExportRowExpand_SSE2(WebPRescaler* const wrk) {
   int x_out;
   uint8_t* const dst = wrk->dst;
@@ -274,8 +245,7 @@ static void RescalerExportRowExpand_SSE2(WebPRescaler* const wrk) {
     for (; x_out < x_out_max; ++x_out) {
       const uint32_t J = frow[x_out];
       const int v = (int)MULT_FIX(J, wrk->fy_scale);
-      assert(v >= 0 && v <= 255);
-      dst[x_out] = v;
+      dst[x_out] = (v > 255) ? 255u : (uint8_t)v;
     }
   } else {
     const uint32_t B = WEBP_RESCALER_FRAC(-wrk->y_accum, wrk->y_sub);
@@ -308,8 +278,7 @@ static void RescalerExportRowExpand_SSE2(WebPRescaler* const wrk) {
                        + (uint64_t)B * irow[x_out];
       const uint32_t J = (uint32_t)((I + ROUNDER) >> WEBP_RESCALER_RFIX);
       const int v = (int)MULT_FIX(J, wrk->fy_scale);
-      assert(v >= 0 && v <= 255);
-      dst[x_out] = v;
+      dst[x_out] = (v > 255) ? 255u : (uint8_t)v;
     }
   }
 }
@@ -328,20 +297,15 @@ static void RescalerExportRowShrink_SSE2(WebPRescaler* const wrk) {
     const int scale_xy = wrk->fxy_scale;
     const __m128i mult_xy = _mm_set_epi32(0, scale_xy, 0, scale_xy);
     const __m128i mult_y = _mm_set_epi32(0, yscale, 0, yscale);
-    const __m128i rounder = _mm_set_epi32(0, ROUNDER, 0, ROUNDER);
     for (x_out = 0; x_out + 8 <= x_out_max; x_out += 8) {
       __m128i A0, A1, A2, A3, B0, B1, B2, B3;
       LoadDispatchAndMult_SSE2(irow + x_out, NULL, &A0, &A1, &A2, &A3);
       LoadDispatchAndMult_SSE2(frow + x_out, &mult_y, &B0, &B1, &B2, &B3);
       {
-        const __m128i C0 = _mm_add_epi64(B0, rounder);
-        const __m128i C1 = _mm_add_epi64(B1, rounder);
-        const __m128i C2 = _mm_add_epi64(B2, rounder);
-        const __m128i C3 = _mm_add_epi64(B3, rounder);
-        const __m128i D0 = _mm_srli_epi64(C0, WEBP_RESCALER_RFIX);   // = frac
-        const __m128i D1 = _mm_srli_epi64(C1, WEBP_RESCALER_RFIX);
-        const __m128i D2 = _mm_srli_epi64(C2, WEBP_RESCALER_RFIX);
-        const __m128i D3 = _mm_srli_epi64(C3, WEBP_RESCALER_RFIX);
+        const __m128i D0 = _mm_srli_epi64(B0, WEBP_RESCALER_RFIX);   // = frac
+        const __m128i D1 = _mm_srli_epi64(B1, WEBP_RESCALER_RFIX);
+        const __m128i D2 = _mm_srli_epi64(B2, WEBP_RESCALER_RFIX);
+        const __m128i D3 = _mm_srli_epi64(B3, WEBP_RESCALER_RFIX);
         const __m128i E0 = _mm_sub_epi64(A0, D0);   // irow[x] - frac
         const __m128i E1 = _mm_sub_epi64(A1, D1);
         const __m128i E2 = _mm_sub_epi64(A2, D2);
@@ -352,14 +316,13 @@ static void RescalerExportRowShrink_SSE2(WebPRescaler* const wrk) {
         const __m128i G1 = _mm_or_si128(D1, F3);
         _mm_storeu_si128((__m128i*)(irow + x_out + 0), G0);
         _mm_storeu_si128((__m128i*)(irow + x_out + 4), G1);
-        ProcessRow_Floor_SSE2(&E0, &E1, &E2, &E3, &mult_xy, dst + x_out);
+        ProcessRow_SSE2(&E0, &E1, &E2, &E3, &mult_xy, dst + x_out);
       }
     }
     for (; x_out < x_out_max; ++x_out) {
-      const uint32_t frac = (int)MULT_FIX(frow[x_out], yscale);
-      const int v = (int)MULT_FIX_FLOOR(irow[x_out] - frac, wrk->fxy_scale);
-      assert(v >= 0 && v <= 255);
-      dst[x_out] = v;
+      const uint32_t frac = (int)MULT_FIX_FLOOR(frow[x_out], yscale);
+      const int v = (int)MULT_FIX(irow[x_out] - frac, wrk->fxy_scale);
+      dst[x_out] = (v > 255) ? 255u : (uint8_t)v;
       irow[x_out] = frac;   // new fractional start
     }
   } else {
@@ -375,8 +338,7 @@ static void RescalerExportRowShrink_SSE2(WebPRescaler* const wrk) {
     }
     for (; x_out < x_out_max; ++x_out) {
       const int v = (int)MULT_FIX(irow[x_out], scale);
-      assert(v >= 0 && v <= 255);
-      dst[x_out] = v;
+      dst[x_out] = (v > 255) ? 255u : (uint8_t)v;
       irow[x_out] = 0;
     }
   }

3rdparty/libwebp/src/enc/backward_references_enc.c

@@ -191,13 +191,14 @@ void VP8LHashChainClear(VP8LHashChain* const p) {
 
 // -----------------------------------------------------------------------------
 
-#define HASH_MULTIPLIER_HI (0xc6a4a793ULL)
-#define HASH_MULTIPLIER_LO (0x5bd1e996ULL)
+static const uint32_t kHashMultiplierHi = 0xc6a4a793u;
+static const uint32_t kHashMultiplierLo = 0x5bd1e996u;
 
-static WEBP_INLINE uint32_t GetPixPairHash64(const uint32_t* const argb) {
+static WEBP_UBSAN_IGNORE_UNSIGNED_OVERFLOW WEBP_INLINE
+uint32_t GetPixPairHash64(const uint32_t* const argb) {
   uint32_t key;
-  key  = (argb[1] * HASH_MULTIPLIER_HI) & 0xffffffffu;
-  key += (argb[0] * HASH_MULTIPLIER_LO) & 0xffffffffu;
+  key  = argb[1] * kHashMultiplierHi;
+  key += argb[0] * kHashMultiplierLo;
   key = key >> (32 - HASH_BITS);
   return key;
 }

3rdparty/libwebp/src/enc/histogram_enc.c

@@ -929,9 +929,8 @@ static int HistogramCombineStochastic(VP8LHistogramSet* const image_histo,
   }
 
   mappings = (int*) WebPSafeMalloc(*num_used, sizeof(*mappings));
-  if (mappings == NULL || !HistoQueueInit(&histo_queue, kHistoQueueSize)) {
-    goto End;
-  }
+  if (mappings == NULL) return 0;
+  if (!HistoQueueInit(&histo_queue, kHistoQueueSize)) goto End;
   // Fill the initial mapping.
   for (j = 0, iter = 0; iter < image_histo->size; ++iter) {
     if (histograms[iter] == NULL) continue;

3rdparty/libwebp/src/enc/predictor_enc.c

@@ -202,7 +202,7 @@ static uint32_t NearLossless(uint32_t value, uint32_t predict,
   }
   if ((value >> 24) == 0 || (value >> 24) == 0xff) {
     // Preserve transparency of fully transparent or fully opaque pixels.
-    a = NearLosslessDiff(value >> 24, predict >> 24);
+    a = NearLosslessDiff((value >> 24) & 0xff, (predict >> 24) & 0xff);
   } else {
     a = NearLosslessComponent(value >> 24, predict >> 24, 0xff, quantization);
   }
@@ -215,12 +215,12 @@ static uint32_t NearLossless(uint32_t value, uint32_t predict,
     // The amount by which green has been adjusted during quantization. It is
     // subtracted from red and blue for compensation, to avoid accumulating two
     // quantization errors in them.
-    green_diff = NearLosslessDiff(new_green, value >> 8);
+    green_diff = NearLosslessDiff(new_green, (value >> 8) & 0xff);
   }
-  r = NearLosslessComponent(NearLosslessDiff(value >> 16, green_diff),
+  r = NearLosslessComponent(NearLosslessDiff((value >> 16) & 0xff, green_diff),
                             (predict >> 16) & 0xff, 0xff - new_green,
                             quantization);
-  b = NearLosslessComponent(NearLosslessDiff(value, green_diff),
+  b = NearLosslessComponent(NearLosslessDiff(value & 0xff, green_diff),
                             predict & 0xff, 0xff - new_green, quantization);
   return ((uint32_t)a << 24) | ((uint32_t)r << 16) | ((uint32_t)g << 8) | b;
 }
@@ -587,7 +587,7 @@ static void GetBestGreenToRed(
       }
     }
   }
-  best_tx->green_to_red_ = green_to_red_best;
+  best_tx->green_to_red_ = (green_to_red_best & 0xff);
 }
 
 static float GetPredictionCostCrossColorBlue(
@@ -666,8 +666,8 @@ static void GetBestGreenRedToBlue(
       break;  // out of iter-loop.
     }
   }
-  best_tx->green_to_blue_ = green_to_blue_best;
-  best_tx->red_to_blue_ = red_to_blue_best;
+  best_tx->green_to_blue_ = green_to_blue_best & 0xff;
+  best_tx->red_to_blue_ = red_to_blue_best & 0xff;
 }
 #undef kGreenRedToBlueMaxIters
 #undef kGreenRedToBlueNumAxis

3rdparty/libwebp/src/enc/quant_enc.c

@@ -33,7 +33,7 @@
 
 // number of non-zero coeffs below which we consider the block very flat
 // (and apply a penalty to complex predictions)
-#define FLATNESS_LIMIT_I16 10      // I16 mode
+#define FLATNESS_LIMIT_I16 0       // I16 mode (special case)
 #define FLATNESS_LIMIT_I4  3       // I4 mode
 #define FLATNESS_LIMIT_UV  2       // UV mode
 #define FLATNESS_PENALTY   140     // roughly ~1bit per block
@@ -988,6 +988,7 @@ static void PickBestIntra16(VP8EncIterator* const it, VP8ModeScore* rd) {
   VP8ModeScore* rd_cur = &rd_tmp;
   VP8ModeScore* rd_best = rd;
   int mode;
+  int is_flat = IsFlatSource16(it->yuv_in_ + Y_OFF_ENC);
 
   rd->mode_i16 = -1;
   for (mode = 0; mode < NUM_PRED_MODES; ++mode) {
@@ -1003,10 +1004,14 @@ static void PickBestIntra16(VP8EncIterator* const it, VP8ModeScore* rd) {
         tlambda ? MULT_8B(tlambda, VP8TDisto16x16(src, tmp_dst, kWeightY)) : 0;
     rd_cur->H = VP8FixedCostsI16[mode];
     rd_cur->R = VP8GetCostLuma16(it, rd_cur);
-    if (mode > 0 &&
-        IsFlat(rd_cur->y_ac_levels[0], kNumBlocks, FLATNESS_LIMIT_I16)) {
-      // penalty to avoid flat area to be mispredicted by complex mode
-      rd_cur->R += FLATNESS_PENALTY * kNumBlocks;
+    if (is_flat) {
+      // refine the first impression (which was in pixel space)
+      is_flat = IsFlat(rd_cur->y_ac_levels[0], kNumBlocks, FLATNESS_LIMIT_I16);
+      if (is_flat) {
+        // Block is very flat. We put emphasis on the distortion being very low!
+        rd_cur->D *= 2;
+        rd_cur->SD *= 2;
+      }
     }
 
     // Since we always examine Intra16 first, we can overwrite *rd directly.
@@ -1087,7 +1092,8 @@ static int PickBestIntra4(VP8EncIterator* const it, VP8ModeScore* const rd) {
                   : 0;
       rd_tmp.H = mode_costs[mode];
 
-      // Add flatness penalty
+      // Add flatness penalty, to avoid flat area to be mispredicted
+      // by a complex mode.
       if (mode > 0 && IsFlat(tmp_levels, kNumBlocks, FLATNESS_LIMIT_I4)) {
         rd_tmp.R = FLATNESS_PENALTY * kNumBlocks;
       } else {
@@ -1242,11 +1248,19 @@ static void RefineUsingDistortion(VP8EncIterator* const it,
       if (mode > 0 && VP8FixedCostsI16[mode] > bit_limit) {
         continue;
       }
+
       if (score < best_score) {
         best_mode = mode;
         best_score = score;
       }
     }
+    if (it->x_ == 0 || it->y_ == 0) {
+      // avoid starting a checkerboard resonance from the border. See bug #432.
+      if (IsFlatSource16(src)) {
+        best_mode = (it->x_ == 0) ? 0 : 2;
+        try_both_modes = 0;  // stick to i16
+      }
+    }
     VP8SetIntra16Mode(it, best_mode);
     // we'll reconstruct later, if i16 mode actually gets selected
   }

3rdparty/libwebp/src/enc/vp8i_enc.h

@@ -32,7 +32,7 @@ extern "C" {
 // version numbers
 #define ENC_MAJ_VERSION 1
 #define ENC_MIN_VERSION 0
-#define ENC_REV_VERSION 2
+#define ENC_REV_VERSION 3
 
 enum { MAX_LF_LEVELS = 64,       // Maximum loop filter level
        MAX_VARIABLE_LEVEL = 67,  // last (inclusive) level with variable cost

3rdparty/libwebp/src/mux/muxi.h

@@ -29,7 +29,7 @@ extern "C" {
 
 #define MUX_MAJ_VERSION 1
 #define MUX_MIN_VERSION 0
-#define MUX_REV_VERSION 2
+#define MUX_REV_VERSION 3
 
 // Chunk object.
 typedef struct WebPChunk WebPChunk;

3rdparty/libwebp/src/utils/bit_reader_inl_utils.h

@@ -104,7 +104,8 @@ void VP8LoadNewBytes(VP8BitReader* const br) {
 }
 
 // Read a bit with proba 'prob'. Speed-critical function!
-static WEBP_INLINE int VP8GetBit(VP8BitReader* const br, int prob) {
+static WEBP_INLINE int VP8GetBit(VP8BitReader* const br,
+                                 int prob, const char label[]) {
   // Don't move this declaration! It makes a big speed difference to store
   // 'range' *before* calling VP8LoadNewBytes(), even if this function doesn't
   // alter br->range_ value.
@@ -129,13 +130,14 @@ static WEBP_INLINE int VP8GetBit(VP8BitReader* const br, int prob) {
       br->bits_ -= shift;
     }
     br->range_ = range - 1;
+    BT_TRACK(br);
     return bit;
   }
 }
 
 // simplified version of VP8GetBit() for prob=0x80 (note shift is always 1 here)
 static WEBP_UBSAN_IGNORE_UNSIGNED_OVERFLOW WEBP_INLINE
-int VP8GetSigned(VP8BitReader* const br, int v) {
+int VP8GetSigned(VP8BitReader* const br, int v, const char label[]) {
   if (br->bits_ < 0) {
     VP8LoadNewBytes(br);
   }
@@ -148,11 +150,13 @@ int VP8GetSigned(VP8BitReader* const br, int v) {
     br->range_ += mask;
     br->range_ |= 1;
     br->value_ -= (bit_t)((split + 1) & mask) << pos;
+    BT_TRACK(br);
     return (v ^ mask) - mask;
   }
 }
 
-static WEBP_INLINE int VP8GetBitAlt(VP8BitReader* const br, int prob) {
+static WEBP_INLINE int VP8GetBitAlt(VP8BitReader* const br,
+                                    int prob, const char label[]) {
   // Don't move this declaration! It makes a big speed difference to store
   // 'range' *before* calling VP8LoadNewBytes(), even if this function doesn't
   // alter br->range_ value.
@@ -179,6 +183,7 @@ static WEBP_INLINE int VP8GetBitAlt(VP8BitReader* const br, int prob) {
       br->bits_ -= shift;
     }
     br->range_ = range;
+    BT_TRACK(br);
     return bit;
   }
 }

3rdparty/libwebp/src/utils/bit_reader_utils.c

@@ -102,17 +102,18 @@ void VP8LoadFinalBytes(VP8BitReader* const br) {
 //------------------------------------------------------------------------------
 // Higher-level calls
 
-uint32_t VP8GetValue(VP8BitReader* const br, int bits) {
+uint32_t VP8GetValue(VP8BitReader* const br, int bits, const char label[]) {
   uint32_t v = 0;
   while (bits-- > 0) {
-    v |= VP8GetBit(br, 0x80) << bits;
+    v |= VP8GetBit(br, 0x80, label) << bits;
   }
   return v;
 }
 
-int32_t VP8GetSignedValue(VP8BitReader* const br, int bits) {
-  const int value = VP8GetValue(br, bits);
-  return VP8Get(br) ? -value : value;
+int32_t VP8GetSignedValue(VP8BitReader* const br, int bits,
+                          const char label[]) {
+  const int value = VP8GetValue(br, bits, label);
+  return VP8Get(br, label) ? -value : value;
 }
 
 //------------------------------------------------------------------------------
@@ -220,3 +221,78 @@ uint32_t VP8LReadBits(VP8LBitReader* const br, int n_bits) {
 }
 
 //------------------------------------------------------------------------------
+// Bit-tracing tool
+
+#if (BITTRACE > 0)
+
+#include <stdlib.h>   // for atexit()
+#include <stdio.h>
+#include <string.h>
+
+#define MAX_NUM_LABELS 32
+static struct {
+  const char* label;
+  int size;
+  int count;
+} kLabels[MAX_NUM_LABELS];
+
+static int last_label = 0;
+static int last_pos = 0;
+static const uint8_t* buf_start = NULL;
+static int init_done = 0;
+
+static void PrintBitTraces(void) {
+  int i;
+  int scale = 1;
+  int total = 0;
+  const char* units = "bits";
+#if (BITTRACE == 2)
+  scale = 8;
+  units = "bytes";
+#endif
+  for (i = 0; i < last_label; ++i) total += kLabels[i].size;
+  if (total < 1) total = 1;   // avoid rounding errors
+  printf("=== Bit traces ===\n");
+  for (i = 0; i < last_label; ++i) {
+    const int skip = 16 - (int)strlen(kLabels[i].label);
+    const int value = (kLabels[i].size + scale - 1) / scale;
+    assert(skip > 0);
+    printf("%s \%*s: %6d %s   \t[%5.2f%%] [count: %7d]\n",
+           kLabels[i].label, skip, "", value, units,
+           100.f * kLabels[i].size / total,
+           kLabels[i].count);
+  }
+  total = (total + scale - 1) / scale;
+  printf("Total: %d %s\n", total, units);
+}
+
+void BitTrace(const struct VP8BitReader* const br, const char label[]) {
+  int i, pos;
+  if (!init_done) {
+    memset(kLabels, 0, sizeof(kLabels));
+    atexit(PrintBitTraces);
+    buf_start = br->buf_;
+    init_done = 1;
+  }
+  pos = (int)(br->buf_ - buf_start) * 8 - br->bits_;
+  // if there's a too large jump, we've changed partition -> reset counter
+  if (abs(pos - last_pos) > 32) {
+    buf_start = br->buf_;
+    pos = 0;
+    last_pos = 0;
+  }
+  if (br->range_ >= 0x7f) pos += kVP8Log2Range[br->range_ - 0x7f];
+  for (i = 0; i < last_label; ++i) {
+    if (!strcmp(label, kLabels[i].label)) break;
+  }
+  if (i == MAX_NUM_LABELS) abort();   // overflow!
+  kLabels[i].label = label;
+  kLabels[i].size += pos - last_pos;
+  kLabels[i].count += 1;
+  if (i == last_label) ++last_label;
+  last_pos = pos;
+}
+
+#endif  // BITTRACE > 0
+
+//------------------------------------------------------------------------------

3rdparty/libwebp/src/utils/bit_reader_utils.h

@@ -21,6 +21,27 @@
 #endif
 #include "src/webp/types.h"
 
+// Warning! This macro triggers quite some MACRO wizardry around func signature!
+#if !defined(BITTRACE)
+#define BITTRACE 0    // 0 = off, 1 = print bits, 2 = print bytes
+#endif
+
+#if (BITTRACE > 0)
+struct VP8BitReader;
+extern void BitTrace(const struct VP8BitReader* const br, const char label[]);
+#define BT_TRACK(br) BitTrace(br, label)
+#define VP8Get(BR, L) VP8GetValue(BR, 1, L)
+#else
+#define BT_TRACK(br)
+// We'll REMOVE the 'const char label[]' from all signatures and calls (!!):
+#define VP8GetValue(BR, N, L) VP8GetValue(BR, N)
+#define VP8Get(BR, L) VP8GetValue(BR, 1, L)
+#define VP8GetSignedValue(BR, N, L) VP8GetSignedValue(BR, N)
+#define VP8GetBit(BR, P, L) VP8GetBit(BR, P)
+#define VP8GetBitAlt(BR, P, L) VP8GetBitAlt(BR, P)
+#define VP8GetSigned(BR, V, L) VP8GetSigned(BR, V)
+#endif
+
 #ifdef __cplusplus
 extern "C" {
 #endif
@@ -92,17 +113,15 @@ void VP8BitReaderSetBuffer(VP8BitReader* const br,
 void VP8RemapBitReader(VP8BitReader* const br, ptrdiff_t offset);
 
 // return the next value made of 'num_bits' bits
-uint32_t VP8GetValue(VP8BitReader* const br, int num_bits);
-static WEBP_INLINE uint32_t VP8Get(VP8BitReader* const br) {
-  return VP8GetValue(br, 1);
-}
+uint32_t VP8GetValue(VP8BitReader* const br, int num_bits, const char label[]);
 
 // return the next value with sign-extension.
-int32_t VP8GetSignedValue(VP8BitReader* const br, int num_bits);
+int32_t VP8GetSignedValue(VP8BitReader* const br, int num_bits,
+                          const char label[]);
 
 // bit_reader_inl.h will implement the following methods:
-//   static WEBP_INLINE int VP8GetBit(VP8BitReader* const br, int prob)
-//   static WEBP_INLINE int VP8GetSigned(VP8BitReader* const br, int v)
+//   static WEBP_INLINE int VP8GetBit(VP8BitReader* const br, int prob, ...)
+//   static WEBP_INLINE int VP8GetSigned(VP8BitReader* const br, int v, ...)
 // and should be included by the .c files that actually need them.
 // This is to avoid recompiling the whole library whenever this file is touched,
 // and also allowing platform-specific ad-hoc hacks.

3rdparty/libwebp/src/utils/bit_writer_utils.c

@@ -70,7 +70,7 @@ static void Flush(VP8BitWriter* const bw) {
       const int value = (bits & 0x100) ? 0x00 : 0xff;
       for (; bw->run_ > 0; --bw->run_) bw->buf_[pos++] = value;
     }
-    bw->buf_[pos++] = bits;
+    bw->buf_[pos++] = bits & 0xff;
     bw->pos_ = pos;
   } else {
     bw->run_++;   // delay writing of bytes 0xff, pending eventual carry.

3rdparty/libwebp/src/utils/color_cache_utils.h

@@ -17,6 +17,7 @@
 
 #include <assert.h>
 
+#include "src/dsp/dsp.h"
 #include "src/webp/types.h"
 
 #ifdef __cplusplus
@@ -30,10 +31,11 @@ typedef struct {
   int hash_bits_;
 } VP8LColorCache;
 
-static const uint64_t kHashMul = 0x1e35a7bdull;
+static const uint32_t kHashMul = 0x1e35a7bdu;
 
-static WEBP_INLINE int VP8LHashPix(uint32_t argb, int shift) {
-  return (int)(((argb * kHashMul) & 0xffffffffu) >> shift);
+static WEBP_UBSAN_IGNORE_UNSIGNED_OVERFLOW WEBP_INLINE
+int VP8LHashPix(uint32_t argb, int shift) {
+  return (int)((argb * kHashMul) >> shift);
 }
 
 static WEBP_INLINE uint32_t VP8LColorCacheLookup(

3rdparty/libwebp/src/utils/huffman_utils.c

@@ -91,7 +91,8 @@ static int BuildHuffmanTable(HuffmanCode* const root_table, int root_bits,
 
   assert(code_lengths_size != 0);
   assert(code_lengths != NULL);
-  assert(root_table != NULL);
+  assert((root_table != NULL && sorted != NULL) ||
+         (root_table == NULL && sorted == NULL));
   assert(root_bits > 0);
 
   // Build histogram of code lengths.
@@ -120,16 +121,22 @@ static int BuildHuffmanTable(HuffmanCode* const root_table, int root_bits,
   for (symbol = 0; symbol < code_lengths_size; ++symbol) {
     const int symbol_code_length = code_lengths[symbol];
     if (code_lengths[symbol] > 0) {
-      sorted[offset[symbol_code_length]++] = symbol;
+      if (sorted != NULL) {
+        sorted[offset[symbol_code_length]++] = symbol;
+      } else {
+        offset[symbol_code_length]++;
+      }
     }
   }
 
   // Special case code with only one value.
   if (offset[MAX_ALLOWED_CODE_LENGTH] == 1) {
-    HuffmanCode code;
-    code.bits = 0;
-    code.value = (uint16_t)sorted[0];
-    ReplicateValue(table, 1, total_size, code);
+    if (sorted != NULL) {
+      HuffmanCode code;
+      code.bits = 0;
+      code.value = (uint16_t)sorted[0];
+      ReplicateValue(table, 1, total_size, code);
+    }
     return total_size;
   }
 
@@ -151,6 +158,7 @@ static int BuildHuffmanTable(HuffmanCode* const root_table, int root_bits,
       if (num_open < 0) {
         return 0;
       }
+      if (root_table == NULL) continue;
       for (; count[len] > 0; --count[len]) {
         HuffmanCode code;
         code.bits = (uint8_t)len;
@@ -169,6 +177,7 @@ static int BuildHuffmanTable(HuffmanCode* const root_table, int root_bits,
       if (num_open < 0) {
         return 0;
       }
+      if (root_table == NULL) continue;
       for (; count[len] > 0; --count[len]) {
         HuffmanCode code;
         if ((key & mask) != low) {
@@ -206,7 +215,10 @@ int VP8LBuildHuffmanTable(HuffmanCode* const root_table, int root_bits,
                           const int code_lengths[], int code_lengths_size) {
   int total_size;
   assert(code_lengths_size <= MAX_CODE_LENGTHS_SIZE);
-  if (code_lengths_size <= SORTED_SIZE_CUTOFF) {
+  if (root_table == NULL) {
+    total_size = BuildHuffmanTable(NULL, root_bits,
+                                   code_lengths, code_lengths_size, NULL);
+  } else if (code_lengths_size <= SORTED_SIZE_CUTOFF) {
     // use local stack-allocated array.
     uint16_t sorted[SORTED_SIZE_CUTOFF];
     total_size = BuildHuffmanTable(root_table, root_bits,

3rdparty/libwebp/src/utils/huffman_utils.h

@@ -78,6 +78,8 @@ void VP8LHtreeGroupsFree(HTreeGroup* const htree_groups);
 // the huffman table.
 // Returns built table size or 0 in case of error (invalid tree or
 // memory error).
+// If root_table is NULL, it returns 0 if a lookup cannot be built, something
+// > 0 otherwise (but not the table size).
 int VP8LBuildHuffmanTable(HuffmanCode* const root_table, int root_bits,
                           const int code_lengths[], int code_lengths_size);
 

3rdparty/libwebp/src/utils/rescaler_utils.c

@@ -84,14 +84,14 @@ int WebPRescalerGetScaledDimensions(int src_width, int src_height,
     int height = *scaled_height;
 
     // if width is unspecified, scale original proportionally to height ratio.
-    if (width == 0) {
+    if (width == 0 && src_height > 0) {
       width =
-          (int)(((uint64_t)src_width * height + src_height / 2) / src_height);
+          (int)(((uint64_t)src_width * height + src_height - 1) / src_height);
     }
     // if height is unspecified, scale original proportionally to width ratio.
-    if (height == 0) {
+    if (height == 0 && src_width > 0) {
       height =
-          (int)(((uint64_t)src_height * width + src_width / 2) / src_width);
+          (int)(((uint64_t)src_height * width + src_width - 1) / src_width);
     }
     // Check if the overall dimensions still make sense.
     if (width <= 0 || height <= 0) {

3rdparty/libwebp/src/utils/thread_utils.c

@@ -217,8 +217,12 @@ static THREADFN ThreadLoop(void* ptr) {
       done = 1;
     }
     // signal to the main thread that we're done (for Sync())
-    pthread_cond_signal(&impl->condition_);
+    // Note the associated mutex does not need to be held when signaling the
+    // condition. Unlocking the mutex first may improve performance in some
+    // implementations, avoiding the case where the waiting thread can't
+    // reacquire the mutex when woken.
     pthread_mutex_unlock(&impl->mutex_);
+    pthread_cond_signal(&impl->condition_);
   }
   return THREAD_RETURN(NULL);    // Thread is finished
 }
@@ -240,7 +244,13 @@ static void ChangeState(WebPWorker* const worker, WebPWorkerStatus new_status) {
     // assign new status and release the working thread if needed
     if (new_status != OK) {
       worker->status_ = new_status;
+      // Note the associated mutex does not need to be held when signaling the
+      // condition. Unlocking the mutex first may improve performance in some
+      // implementations, avoiding the case where the waiting thread can't
+      // reacquire the mutex when woken.
+      pthread_mutex_unlock(&impl->mutex_);
       pthread_cond_signal(&impl->condition_);
+      return;
     }
   }
   pthread_mutex_unlock(&impl->mutex_);

3rdparty/libwebp/src/utils/utils.h

@@ -92,14 +92,14 @@ static WEBP_INLINE uint32_t GetLE32(const uint8_t* const data) {
 // Store 16, 24 or 32 bits in little-endian order.
 static WEBP_INLINE void PutLE16(uint8_t* const data, int val) {
   assert(val < (1 << 16));
-  data[0] = (val >> 0);
-  data[1] = (val >> 8);
+  data[0] = (val >> 0) & 0xff;
+  data[1] = (val >> 8) & 0xff;
 }
 
 static WEBP_INLINE void PutLE24(uint8_t* const data, int val) {
   assert(val < (1 << 24));
   PutLE16(data, val & 0xffff);
-  data[2] = (val >> 16);
+  data[2] = (val >> 16) & 0xff;
 }
 
 static WEBP_INLINE void PutLE32(uint8_t* const data, uint32_t val) {

3rdparty/libwebp/src/webp/encode.h

@@ -62,6 +62,10 @@ WEBP_EXTERN size_t WebPEncodeBGRA(const uint8_t* bgra,
 // These functions are the equivalent of the above, but compressing in a
 // lossless manner. Files are usually larger than lossy format, but will
 // not suffer any compression loss.
+// Note these functions, like the lossy versions, use the library's default
+// settings. For lossless this means 'exact' is disabled. RGB values in
+// transparent areas will be modified to improve compression. To avoid this,
+// use WebPEncode() and set WebPConfig::exact to 1.
 WEBP_EXTERN size_t WebPEncodeLosslessRGB(const uint8_t* rgb,
                                          int width, int height, int stride,
                                          uint8_t** output);

3rdparty/openexr/IlmImf/ImfSimd.h

@@ -45,7 +45,7 @@
 
 
 // GCC and Visual Studio SSE2 compiler flags
-#if defined __SSE2__ || (_MSC_VER >= 1300 && !_M_CEE_PURE)
+#if defined __SSE2__ || (defined(_M_X64) || _M_IX86_FP == 2)
     #define IMF_HAVE_SSE2 1
 #endif
 

3rdparty/openvx/hal/CMakeLists.txt

@@ -5,7 +5,7 @@ target_include_directories(openvx_hal PUBLIC
   ${CMAKE_SOURCE_DIR}/modules/core/include
   ${CMAKE_SOURCE_DIR}/modules/imgproc/include
   ${OPENVX_INCLUDE_DIR})
-target_link_libraries(openvx_hal LINK_PUBLIC ${OPENVX_LIBRARIES})
+target_link_libraries(openvx_hal PUBLIC ${OPENVX_LIBRARIES})
 set_target_properties(openvx_hal PROPERTIES ARCHIVE_OUTPUT_DIRECTORY ${3P_LIBRARY_OUTPUT_PATH})
 if(NOT BUILD_SHARED_LIBS)
   ocv_install_target(openvx_hal EXPORT OpenCVModules ARCHIVE DESTINATION ${OPENCV_3P_LIB_INSTALL_PATH} COMPONENT dev)

3rdparty/protobuf/CMakeLists.txt

@@ -22,6 +22,7 @@ else()
                                        -Wenum-compare-switch
                                        -Wsuggest-override -Winconsistent-missing-override
                                        -Wimplicit-fallthrough
+                                       -Warray-bounds  # GCC 9+
   )
 endif()
 if(CV_ICC)

3rdparty/tbb/CMakeLists.txt

@@ -5,10 +5,11 @@ if (WIN32 AND NOT ARM)
   message(FATAL_ERROR "BUILD_TBB option supports Windows on ARM only!\nUse regular official TBB build instead of the BUILD_TBB option!")
 endif()
 
-ocv_update(OPENCV_TBB_RELEASE "2019_U8")
-ocv_update(OPENCV_TBB_RELEASE_MD5 "7c371d0f62726154d2c568a85697a0ad")
+ocv_update(OPENCV_TBB_RELEASE "v2020.0")
+ocv_update(OPENCV_TBB_RELEASE_MD5 "5858dd01ec007c139d5d178b21e06dae")
 ocv_update(OPENCV_TBB_FILENAME "${OPENCV_TBB_RELEASE}.tar.gz")
-ocv_update(OPENCV_TBB_SUBDIR "tbb-${OPENCV_TBB_RELEASE}")
+string(REGEX REPLACE "^v" "" OPENCV_TBB_RELEASE_ "${OPENCV_TBB_RELEASE}")
+ocv_update(OPENCV_TBB_SUBDIR "tbb-${OPENCV_TBB_RELEASE_}")
 
 set(tbb_src_dir "${OpenCV_BINARY_DIR}/3rdparty/tbb")
 ocv_download(FILENAME ${OPENCV_TBB_FILENAME}
@@ -34,10 +35,12 @@ ocv_include_directories("${tbb_src_dir}/include"
 file(GLOB lib_srcs "${tbb_src_dir}/src/tbb/*.cpp")
 file(GLOB lib_hdrs "${tbb_src_dir}/src/tbb/*.h")
 list(APPEND lib_srcs "${tbb_src_dir}/src/rml/client/rml_tbb.cpp")
+ocv_list_filterout(lib_srcs "${tbb_src_dir}/src/tbb/tbbbind.cpp")  # hwloc.h requirement
 
 if (WIN32)
   add_definitions(/D__TBB_DYNAMIC_LOAD_ENABLED=0
                   /D__TBB_BUILD=1
+                  /DTBB_SUPPRESS_DEPRECATED_MESSAGES=1
                   /DTBB_NO_LEGACY=1
                   /D_UNICODE
                   /DUNICODE

CMakeLists.txt

@@ -29,40 +29,19 @@ else()
   cmake_minimum_required(VERSION "${MIN_VER_CMAKE}" FATAL_ERROR)
 endif()
 
-option(ENABLE_PIC "Generate position independent code (necessary for shared libraries)" TRUE)
-set(CMAKE_POSITION_INDEPENDENT_CODE ${ENABLE_PIC})
-
-set(OPENCV_MATHJAX_RELPATH "https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.0" CACHE STRING "URI to a MathJax installation")
-
-# Following block can break build in case of cross-compiling
-# but CMAKE_CROSSCOMPILING variable will be set only on project(OpenCV) command
-# so we will try to detect cross-compiling by the presence of CMAKE_TOOLCHAIN_FILE
-if(NOT DEFINED CMAKE_INSTALL_PREFIX)
-  if(NOT CMAKE_TOOLCHAIN_FILE)
-    # it _must_ go before project(OpenCV) in order to work
-    if(WIN32)
-      set(CMAKE_INSTALL_PREFIX "${CMAKE_BINARY_DIR}/install" CACHE PATH "Installation Directory")
-    else()
-      set(CMAKE_INSTALL_PREFIX "/usr/local" CACHE PATH "Installation Directory")
-    endif()
-  else()
-    #Android: set output folder to ${CMAKE_BINARY_DIR}
-    set(LIBRARY_OUTPUT_PATH_ROOT ${CMAKE_BINARY_DIR} CACHE PATH "root for library output, set this to change where android libs are compiled to" )
-    # any cross-compiling
-    set(CMAKE_INSTALL_PREFIX "${CMAKE_BINARY_DIR}/install" CACHE PATH "Installation Directory")
-  endif()
-endif()
-
+#
+# Configure CMake policies
+#
 if(POLICY CMP0026)
   cmake_policy(SET CMP0026 NEW)
 endif()
 
 if(POLICY CMP0042)
-  cmake_policy(SET CMP0042 NEW)
+  cmake_policy(SET CMP0042 NEW)  # CMake 3.0+ (2.8.12): MacOS "@rpath" in target's install name
 endif()
 
 if(POLICY CMP0046)
-  cmake_policy(SET CMP0046 NEW)
+  cmake_policy(SET CMP0046 NEW)  # warn about non-existed dependencies
 endif()
 
 if(POLICY CMP0051)
@@ -74,17 +53,25 @@ if(POLICY CMP0054)  # CMake 3.1: Only interpret if() arguments as variables or k
 endif()
 
 if(POLICY CMP0056)
-  cmake_policy(SET CMP0056 NEW)
+  cmake_policy(SET CMP0056 NEW)  # try_compile(): link flags
 endif()
 
 if(POLICY CMP0067)
-  cmake_policy(SET CMP0067 NEW)
+  cmake_policy(SET CMP0067 NEW)  # CMake 3.8: try_compile(): honor language standard variables (like C++11)
 endif()
 
 if(POLICY CMP0068)
   cmake_policy(SET CMP0068 NEW)  # CMake 3.9+: `RPATH` settings on macOS do not affect `install_name`.
 endif()
 
+if(POLICY CMP0075)
+  cmake_policy(SET CMP0075 NEW)  # CMake 3.12+: Include file check macros honor `CMAKE_REQUIRED_LIBRARIES`
+endif()
+
+
+#
+# Configure OpenCV CMake hooks
+#
 include(cmake/OpenCVUtils.cmake)
 ocv_cmake_reset_hooks()
 ocv_check_environment_variables(OPENCV_CMAKE_HOOKS_DIR)
@@ -97,49 +84,46 @@ endif()
 
 ocv_cmake_hook(CMAKE_INIT)
 
-# must go before the project command
+# must go before the project()/enable_language() commands
 ocv_update(CMAKE_CONFIGURATION_TYPES "Debug;Release" CACHE STRING "Configs" FORCE)
 if(DEFINED CMAKE_BUILD_TYPE)
-  set_property( CACHE CMAKE_BUILD_TYPE PROPERTY STRINGS ${CMAKE_CONFIGURATION_TYPES} )
+  set_property(CACHE CMAKE_BUILD_TYPE PROPERTY STRINGS "${CMAKE_CONFIGURATION_TYPES}")
+endif()
+
+option(ENABLE_PIC "Generate position independent code (necessary for shared libraries)" TRUE)
+set(CMAKE_POSITION_INDEPENDENT_CODE ${ENABLE_PIC})
+
+ocv_cmake_hook(PRE_CMAKE_BOOTSTRAP)
+
+# Bootstap CMake system: setup CMAKE_SYSTEM_NAME and other vars
+enable_language(CXX C)
+
+ocv_cmake_hook(POST_CMAKE_BOOTSTRAP)
+
+if(NOT OPENCV_SKIP_CMAKE_SYSTEM_FILE)
+  include("cmake/platforms/OpenCV-${CMAKE_SYSTEM_NAME}.cmake" OPTIONAL RESULT_VARIABLE "OPENCV_CMAKE_SYSTEM_FILE")
+  if(NOT OPENCV_CMAKE_SYSTEM_FILE)
+    message(STATUS "OpenCV: system-specific configuration file is not found: '${CMAKE_SYSTEM_NAME}'")
+  endif()
+endif()
+
+if(CMAKE_INSTALL_PREFIX_INITIALIZED_TO_DEFAULT)  # https://cmake.org/cmake/help/latest/variable/CMAKE_INSTALL_PREFIX_INITIALIZED_TO_DEFAULT.html
+  if(NOT CMAKE_TOOLCHAIN_FILE)
+    if(WIN32)
+      set(CMAKE_INSTALL_PREFIX "${CMAKE_BINARY_DIR}/install" CACHE PATH "Installation Directory" FORCE)
+    else()
+      set(CMAKE_INSTALL_PREFIX "/usr/local" CACHE PATH "Installation Directory" FORCE)
+    endif()
+  else()
+    # any cross-compiling
+    set(CMAKE_INSTALL_PREFIX "${CMAKE_BINARY_DIR}/install" CACHE PATH "Installation Directory" FORCE)
+  endif()
 endif()
 
 enable_testing()
 
 project(OpenCV CXX C)
 
-if(CMAKE_SYSTEM_NAME MATCHES WindowsPhone OR CMAKE_SYSTEM_NAME MATCHES WindowsStore)
-  set(WINRT TRUE)
-endif()
-
-if(WINRT OR WINCE)
-  add_definitions(-DNO_GETENV)
-endif()
-
-if(WINRT)
-  add_definitions(-DWINRT)
-
-  # Making definitions available to other configurations and
-  # to filter dependency restrictions at compile time.
-  if(CMAKE_SYSTEM_NAME MATCHES WindowsPhone)
-    set(WINRT_PHONE TRUE)
-    add_definitions(-DWINRT_PHONE)
-  elseif(CMAKE_SYSTEM_NAME MATCHES WindowsStore)
-    set(WINRT_STORE TRUE)
-    add_definitions(-DWINRT_STORE)
-  endif()
-
-  if(CMAKE_SYSTEM_VERSION MATCHES 10)
-    set(WINRT_10 TRUE)
-    add_definitions(-DWINRT_10)
-  elseif(CMAKE_SYSTEM_VERSION MATCHES 8.1)
-    set(WINRT_8_1 TRUE)
-    add_definitions(-DWINRT_8_1)
-  elseif(CMAKE_SYSTEM_VERSION MATCHES 8.0)
-    set(WINRT_8_0 TRUE)
-    add_definitions(-DWINRT_8_0)
-  endif()
-endif()
-
 if(MSVC)
   set(CMAKE_USE_RELATIVE_PATHS ON CACHE INTERNAL "" FORCE)
 endif()
@@ -291,6 +275,9 @@ OCV_OPTION(WITH_VULKAN "Include Vulkan support" OFF
 OCV_OPTION(WITH_INF_ENGINE "Include Intel Inference Engine support" OFF
   VISIBLE_IF TRUE
   VERIFY INF_ENGINE_TARGET)
+OCV_OPTION(WITH_NGRAPH "Include nGraph support" WITH_INF_ENGINE
+  VISIBLE_IF TRUE
+  VERIFY TARGET ngraph::ngraph)
 OCV_OPTION(WITH_JASPER "Include JPEG2K support" ON
   VISIBLE_IF NOT IOS
   VERIFY HAVE_JASPER)
@@ -465,7 +452,7 @@ OCV_OPTION(INSTALL_TESTS            "Install accuracy and performance test binar
 # OpenCV build options
 # ===================================================
 OCV_OPTION(ENABLE_CCACHE              "Use ccache"                                               (UNIX AND NOT IOS AND (CMAKE_GENERATOR MATCHES "Makefile" OR CMAKE_GENERATOR MATCHES "Ninja")) )
-OCV_OPTION(ENABLE_PRECOMPILED_HEADERS "Use precompiled headers"                                  ON IF (MSVC OR (NOT IOS AND NOT CMAKE_CROSSCOMPILING) ) )
+OCV_OPTION(ENABLE_PRECOMPILED_HEADERS "Use precompiled headers"                                  MSVC IF (MSVC OR (NOT IOS AND NOT CMAKE_CROSSCOMPILING) ) )
 OCV_OPTION(ENABLE_SOLUTION_FOLDERS    "Solution folder in Visual Studio or in other IDEs"        (MSVC_IDE OR CMAKE_GENERATOR MATCHES Xcode) )
 OCV_OPTION(ENABLE_PROFILING           "Enable profiling in the GCC compiler (Add flags: -g -pg)" OFF  IF CV_GCC )
 OCV_OPTION(ENABLE_COVERAGE            "Enable coverage collection with  GCov"                    OFF  IF CV_GCC )
@@ -493,6 +480,7 @@ OCV_OPTION(CV_DISABLE_OPTIMIZATION    "Disable explicit optimized code (dispatch
 OCV_OPTION(CV_TRACE                   "Enable OpenCV code trace" ON)
 OCV_OPTION(OPENCV_GENERATE_SETUPVARS  "Generate setup_vars* scripts" ON IF (NOT ANDROID AND NOT APPLE_FRAMEWORK) )
 OCV_OPTION(ENABLE_CONFIG_VERIFICATION "Fail build if actual configuration doesn't match requested (WITH_XXX != HAVE_XXX)" OFF)
+OCV_OPTION(OPENCV_ENABLE_MEMALIGN     "Enable posix_memalign or memalign usage" ON)
 
 OCV_OPTION(ENABLE_PYLINT              "Add target with Pylint checks"                            (BUILD_DOCS OR BUILD_EXAMPLES) IF (NOT CMAKE_CROSSCOMPILING AND NOT APPLE_FRAMEWORK) )
 OCV_OPTION(ENABLE_FLAKE8              "Add target with Python flake8 checker"                    (BUILD_DOCS OR BUILD_EXAMPLES) IF (NOT CMAKE_CROSSCOMPILING AND NOT APPLE_FRAMEWORK) )
@@ -501,6 +489,7 @@ if(ENABLE_IMPL_COLLECTION)
   add_definitions(-DCV_COLLECT_IMPL_DATA)
 endif()
 
+set(OPENCV_MATHJAX_RELPATH "https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.0" CACHE STRING "URI to a MathJax installation")
 
 # ----------------------------------------------------------------------------
 #  Get actual OpenCV version number from sources
@@ -513,9 +502,21 @@ ocv_cmake_hook(POST_OPTIONS)
 #  Build & install layouts
 # ----------------------------------------------------------------------------
 
+if(OPENCV_TEST_DATA_PATH)
+  get_filename_component(OPENCV_TEST_DATA_PATH ${OPENCV_TEST_DATA_PATH} ABSOLUTE)
+endif()
+
 # Save libs and executables in the same place
 set(EXECUTABLE_OUTPUT_PATH "${CMAKE_BINARY_DIR}/bin" CACHE PATH "Output directory for applications")
 
+if(ANDROID)
+  set(LIBRARY_OUTPUT_PATH                "${OpenCV_BINARY_DIR}/lib/${ANDROID_NDK_ABI_NAME}")
+  ocv_update(3P_LIBRARY_OUTPUT_PATH      "${OpenCV_BINARY_DIR}/3rdparty/lib/${ANDROID_NDK_ABI_NAME}")
+else()
+  set(LIBRARY_OUTPUT_PATH                "${OpenCV_BINARY_DIR}/lib")
+  ocv_update(3P_LIBRARY_OUTPUT_PATH      "${OpenCV_BINARY_DIR}/3rdparty/lib")
+endif()
+
 if(ANDROID)
   if(ANDROID_ABI MATCHES "NEON")
     set(ENABLE_NEON ON)
@@ -525,134 +526,14 @@ if(ANDROID)
   endif()
 endif()
 
-if(ANDROID OR WIN32)
-  ocv_update(OPENCV_DOC_INSTALL_PATH doc)
-endif()
-
-if(WIN32 AND CMAKE_HOST_SYSTEM_NAME MATCHES Windows)
-  if(DEFINED OpenCV_RUNTIME AND DEFINED OpenCV_ARCH)
-    ocv_update(OpenCV_INSTALL_BINARIES_PREFIX "${OpenCV_ARCH}/${OpenCV_RUNTIME}/")
-  else()
-    message(STATUS "Can't detect runtime and/or arch")
-    ocv_update(OpenCV_INSTALL_BINARIES_PREFIX "")
-  endif()
-elseif(ANDROID)
-  ocv_update(OpenCV_INSTALL_BINARIES_PREFIX "sdk/native/")
-else()
-  ocv_update(OpenCV_INSTALL_BINARIES_PREFIX "")
-endif()
-
-if(ANDROID)
-  ocv_update(OPENCV_SAMPLES_BIN_INSTALL_PATH "${OpenCV_INSTALL_BINARIES_PREFIX}samples/${ANDROID_NDK_ABI_NAME}")
-else()
-  ocv_update(OPENCV_SAMPLES_BIN_INSTALL_PATH "${OpenCV_INSTALL_BINARIES_PREFIX}samples")
-endif()
-
-if(ANDROID)
-  ocv_update(OPENCV_BIN_INSTALL_PATH "${OpenCV_INSTALL_BINARIES_PREFIX}bin/${ANDROID_NDK_ABI_NAME}")
-else()
-  ocv_update(OPENCV_BIN_INSTALL_PATH "${OpenCV_INSTALL_BINARIES_PREFIX}bin")
-endif()
-
-if(NOT OPENCV_TEST_INSTALL_PATH)
-  ocv_update(OPENCV_TEST_INSTALL_PATH "${OPENCV_BIN_INSTALL_PATH}")
-endif()
-
-if (OPENCV_TEST_DATA_PATH)
-  get_filename_component(OPENCV_TEST_DATA_PATH ${OPENCV_TEST_DATA_PATH} ABSOLUTE)
-endif()
-
-if(ANDROID)
-  ocv_update(OPENCV_TEST_DATA_INSTALL_PATH "sdk/etc/testdata")
-elseif(WIN32)
-  ocv_update(OPENCV_TEST_DATA_INSTALL_PATH "testdata")
-endif()
-
-if(ANDROID)
-  set(LIBRARY_OUTPUT_PATH                "${OpenCV_BINARY_DIR}/lib/${ANDROID_NDK_ABI_NAME}")
-  ocv_update(3P_LIBRARY_OUTPUT_PATH      "${OpenCV_BINARY_DIR}/3rdparty/lib/${ANDROID_NDK_ABI_NAME}")
-  ocv_update(OPENCV_LIB_INSTALL_PATH     sdk/native/libs/${ANDROID_NDK_ABI_NAME})
-  ocv_update(OPENCV_LIB_ARCHIVE_INSTALL_PATH sdk/native/staticlibs/${ANDROID_NDK_ABI_NAME})
-  ocv_update(OPENCV_3P_LIB_INSTALL_PATH  sdk/native/3rdparty/libs/${ANDROID_NDK_ABI_NAME})
-  ocv_update(OPENCV_CONFIG_INSTALL_PATH  sdk/native/jni)
-  ocv_update(OPENCV_INCLUDE_INSTALL_PATH sdk/native/jni/include)
-  ocv_update(OPENCV_SAMPLES_SRC_INSTALL_PATH samples/native)
-  ocv_update(OPENCV_OTHER_INSTALL_PATH   sdk/etc)
-  ocv_update(OPENCV_LICENSES_INSTALL_PATH "${OPENCV_OTHER_INSTALL_PATH}/licenses")
-else()
-  set(LIBRARY_OUTPUT_PATH                "${OpenCV_BINARY_DIR}/lib")
-  ocv_update(3P_LIBRARY_OUTPUT_PATH      "${OpenCV_BINARY_DIR}/3rdparty/lib${LIB_SUFFIX}")
-
-  if(WIN32 AND CMAKE_HOST_SYSTEM_NAME MATCHES Windows)
-    if(OpenCV_STATIC)
-      ocv_update(OPENCV_LIB_INSTALL_PATH   "${OpenCV_INSTALL_BINARIES_PREFIX}staticlib${LIB_SUFFIX}")
-    else()
-      ocv_update(OPENCV_LIB_INSTALL_PATH   "${OpenCV_INSTALL_BINARIES_PREFIX}lib${LIB_SUFFIX}")
-    endif()
-    ocv_update(OPENCV_3P_LIB_INSTALL_PATH  "${OpenCV_INSTALL_BINARIES_PREFIX}staticlib${LIB_SUFFIX}")
-    ocv_update(OPENCV_SAMPLES_SRC_INSTALL_PATH    samples)
-    ocv_update(OPENCV_JAR_INSTALL_PATH java)
-    ocv_update(OPENCV_OTHER_INSTALL_PATH   etc)
-    ocv_update(OPENCV_CONFIG_INSTALL_PATH  ".")
-    ocv_update(OPENCV_INCLUDE_INSTALL_PATH "include")
-    ocv_update(OPENCV_LICENSES_INSTALL_PATH "${OPENCV_OTHER_INSTALL_PATH}/licenses")
-  else()
-    # Note: layout differs from OpenCV 3.4
-    include(GNUInstallDirs)
-    ocv_update(OPENCV_INCLUDE_INSTALL_PATH       "${CMAKE_INSTALL_INCLUDEDIR}/opencv4")
-    ocv_update(OPENCV_LIB_INSTALL_PATH           "${CMAKE_INSTALL_LIBDIR}")
-    ocv_update(OPENCV_CONFIG_INSTALL_PATH        "${OPENCV_LIB_INSTALL_PATH}/cmake/opencv4")
-    ocv_update(OPENCV_3P_LIB_INSTALL_PATH        "${OPENCV_LIB_INSTALL_PATH}/opencv4/3rdparty")
-    ocv_update(OPENCV_SAMPLES_SRC_INSTALL_PATH   "${CMAKE_INSTALL_DATAROOTDIR}/opencv4/samples")
-    ocv_update(OPENCV_DOC_INSTALL_PATH           "${CMAKE_INSTALL_DATAROOTDIR}/doc/opencv4")
-    ocv_update(OPENCV_JAR_INSTALL_PATH           "${CMAKE_INSTALL_DATAROOTDIR}/java/opencv4")
-    ocv_update(OPENCV_TEST_DATA_INSTALL_PATH     "${CMAKE_INSTALL_DATAROOTDIR}/opencv4/testdata")
-    ocv_update(OPENCV_OTHER_INSTALL_PATH         "${CMAKE_INSTALL_DATAROOTDIR}/opencv4")
-    ocv_update(OPENCV_LICENSES_INSTALL_PATH      "${CMAKE_INSTALL_DATAROOTDIR}/licenses/opencv4")
-  endif()
-  #ocv_update(OPENCV_PYTHON_INSTALL_PATH  "python")  # no default value, see https://github.com/opencv/opencv/issues/13202
-endif()
-
-ocv_update(CMAKE_INSTALL_RPATH "${CMAKE_INSTALL_PREFIX}/${OPENCV_LIB_INSTALL_PATH}")
-set(CMAKE_INSTALL_RPATH_USE_LINK_PATH TRUE)
-
-if(INSTALL_TO_MANGLED_PATHS)
-  foreach(v
-      OPENCV_INCLUDE_INSTALL_PATH
-      # file names include version (.so/.dll): OPENCV_LIB_INSTALL_PATH
-      OPENCV_CONFIG_INSTALL_PATH
-      OPENCV_3P_LIB_INSTALL_PATH
-      OPENCV_SAMPLES_SRC_INSTALL_PATH
-      OPENCV_DOC_INSTALL_PATH
-      # JAR file name includes version: OPENCV_JAR_INSTALL_PATH
-      OPENCV_TEST_DATA_INSTALL_PATH
-      OPENCV_OTHER_INSTALL_PATH
-    )
-    string(REGEX REPLACE "opencv[0-9]*" "opencv-${OPENCV_VERSION}" ${v} "${${v}}")
-  endforeach()
-endif()
-
-if(ANDROID)
-  ocv_update(OPENCV_JNI_INSTALL_PATH "${OPENCV_LIB_INSTALL_PATH}")
-elseif(INSTALL_CREATE_DISTRIB)
-  ocv_update(OPENCV_JNI_INSTALL_PATH "${OPENCV_JAR_INSTALL_PATH}/${OpenCV_ARCH}")
-else()
-  ocv_update(OPENCV_JNI_INSTALL_PATH "${OPENCV_JAR_INSTALL_PATH}")
-endif()
-ocv_update(OPENCV_JNI_BIN_INSTALL_PATH "${OPENCV_JNI_INSTALL_PATH}")
-
-if(NOT OPENCV_LIB_ARCHIVE_INSTALL_PATH)
-  set(OPENCV_LIB_ARCHIVE_INSTALL_PATH ${OPENCV_LIB_INSTALL_PATH})
-endif()
-
 if(WIN32)
   # Postfix of DLLs:
-  set(OPENCV_DLLVERSION "${OPENCV_VERSION_MAJOR}${OPENCV_VERSION_MINOR}${OPENCV_VERSION_PATCH}")
-  set(OPENCV_DEBUG_POSTFIX d)
+  ocv_update(OPENCV_DLLVERSION "${OPENCV_VERSION_MAJOR}${OPENCV_VERSION_MINOR}${OPENCV_VERSION_PATCH}")
+  ocv_update(OPENCV_DEBUG_POSTFIX d)
 else()
   # Postfix of so's:
-  set(OPENCV_DLLVERSION "")
-  set(OPENCV_DEBUG_POSTFIX "")
+  ocv_update(OPENCV_DLLVERSION "")
+  ocv_update(OPENCV_DEBUG_POSTFIX "")
 endif()
 
 if(DEFINED CMAKE_DEBUG_POSTFIX)
@@ -663,6 +544,8 @@ if((INSTALL_CREATE_DISTRIB AND BUILD_SHARED_LIBS AND NOT DEFINED BUILD_opencv_wo
   set(BUILD_opencv_world ON CACHE INTERNAL "")
 endif()
 
+include(cmake/OpenCVInstallLayout.cmake)
+
 # ----------------------------------------------------------------------------
 #  Path for build/platform -specific headers
 # ----------------------------------------------------------------------------
@@ -746,10 +629,15 @@ if(UNIX)
     set(HAVE_PTHREAD 1)
   endif()
 
-  CHECK_SYMBOL_EXISTS(posix_memalign stdlib.h HAVE_POSIX_MEMALIGN)
-  CHECK_INCLUDE_FILE(malloc.h HAVE_MALLOC_H)
-  if(HAVE_MALLOC_H)
-    CHECK_SYMBOL_EXISTS(memalign malloc.h HAVE_MEMALIGN)
+  if(OPENCV_ENABLE_MEMALIGN)
+    CHECK_SYMBOL_EXISTS(posix_memalign stdlib.h HAVE_POSIX_MEMALIGN)
+    CHECK_INCLUDE_FILE(malloc.h HAVE_MALLOC_H)
+    if(HAVE_MALLOC_H)
+      CHECK_SYMBOL_EXISTS(memalign malloc.h HAVE_MEMALIGN)
+    endif()
+    # TODO:
+    # - _aligned_malloc() on Win32
+    # - std::aligned_alloc() C++17 / C11
   endif()
 endif()
 
@@ -1521,26 +1409,54 @@ endif()
 
 if(WITH_INF_ENGINE OR INF_ENGINE_TARGET)
   if(INF_ENGINE_TARGET)
+    list(GET INF_ENGINE_TARGET 0 ie_target)
     set(__msg "YES (${INF_ENGINE_RELEASE} / ${INF_ENGINE_VERSION})")
-    get_target_property(_lib ${INF_ENGINE_TARGET} IMPORTED_LOCATION)
-    get_target_property(_lib_imp_rel ${INF_ENGINE_TARGET} IMPORTED_IMPLIB_RELEASE)
-    get_target_property(_lib_imp_dbg ${INF_ENGINE_TARGET} IMPORTED_IMPLIB_DEBUG)
-    get_target_property(_lib_rel ${INF_ENGINE_TARGET} IMPORTED_LOCATION_RELEASE)
-    get_target_property(_lib_dbg ${INF_ENGINE_TARGET} IMPORTED_LOCATION_DEBUG)
+    get_target_property(_lib ${ie_target} IMPORTED_LOCATION)
+    get_target_property(_lib_imp_rel ${ie_target} IMPORTED_IMPLIB_RELEASE)
+    get_target_property(_lib_imp_dbg ${ie_target} IMPORTED_IMPLIB_DEBUG)
+    get_target_property(_lib_rel ${ie_target} IMPORTED_LOCATION_RELEASE)
+    get_target_property(_lib_dbg ${ie_target} IMPORTED_LOCATION_DEBUG)
     ocv_build_features_string(_lib
       IF _lib THEN "${_lib}"
       IF _lib_imp_rel AND _lib_imp_dbg THEN "${_lib_imp_rel} / ${_lib_imp_dbg}"
       IF _lib_rel AND _lib_dbg THEN "${_lib_rel} / ${_lib_dbg}"
+      IF _lib_rel  THEN "${_lib_rel}"
+      IF _lib_dbg  THEN "${_lib_dbg}"
       ELSE "unknown"
     )
-    get_target_property(_inc ${INF_ENGINE_TARGET} INTERFACE_INCLUDE_DIRECTORIES)
+    get_target_property(_inc ${ie_target} INTERFACE_INCLUDE_DIRECTORIES)
     status("    Inference Engine:" "${__msg}")
-    status("                libs:" "${_lib}")
-    status("            includes:" "${_inc}")
+    status("        * libs:" "${_lib}")
+    status("        * includes:" "${_inc}")
   else()
     status("    Inference Engine:"     "NO")
   endif()
 endif()
+if(WITH_NGRAPH OR HAVE_NGRAPH)
+  if(HAVE_NGRAPH)
+    set(__target ngraph::ngraph)
+    set(__msg "YES (${ngraph_VERSION})")
+    get_target_property(_lib ${__target} IMPORTED_LOCATION)
+    get_target_property(_lib_imp_rel ${__target} IMPORTED_IMPLIB_RELEASE)
+    get_target_property(_lib_imp_dbg ${__target} IMPORTED_IMPLIB_DEBUG)
+    get_target_property(_lib_rel ${__target} IMPORTED_LOCATION_RELEASE)
+    get_target_property(_lib_dbg ${__target} IMPORTED_LOCATION_DEBUG)
+    ocv_build_features_string(_lib
+      IF _lib THEN "${_lib}"
+      IF _lib_imp_rel AND _lib_imp_dbg THEN "${_lib_imp_rel} / ${_lib_imp_dbg}"
+      IF _lib_rel AND _lib_dbg THEN "${_lib_rel} / ${_lib_dbg}"
+      IF _lib_rel  THEN "${_lib_rel}"
+      IF _lib_dbg  THEN "${_lib_dbg}"
+      ELSE "unknown"
+    )
+    get_target_property(_inc ${__target} INTERFACE_INCLUDE_DIRECTORIES)
+    status("    nGraph:" "${__msg}")
+    status("        * libs:" "${_lib}")
+    status("        * includes:" "${_inc}")
+  else()
+    status("    nGraph:"     "NO")
+  endif()
+endif()
 
 if(WITH_EIGEN OR HAVE_EIGEN)
   status("    Eigen:"      HAVE_EIGEN       THEN "YES (ver ${EIGEN_WORLD_VERSION}.${EIGEN_MAJOR_VERSION}.${EIGEN_MINOR_VERSION})" ELSE NO)
@@ -1637,7 +1553,7 @@ if(FLAKE8_FOUND AND FLAKE8_EXECUTABLE)
 endif()
 
 # ========================== java ==========================
-if(BUILD_JAVA OR BUILD_opencv_java)
+if(BUILD_JAVA)
   status("")
   status("  Java:"            BUILD_FAT_JAVA_LIB  THEN "export all functions"                                      ELSE "")
   status("    ant:"           ANT_EXECUTABLE      THEN "${ANT_EXECUTABLE} (ver ${ANT_VERSION})"                    ELSE NO)

LICENSE

@@ -13,6 +13,7 @@ Copyright (C) 2009-2016, NVIDIA Corporation, all rights reserved.
 Copyright (C) 2010-2013, Advanced Micro Devices, Inc., all rights reserved.
 Copyright (C) 2015-2016, OpenCV Foundation, all rights reserved.
 Copyright (C) 2015-2016, Itseez Inc., all rights reserved.
+Copyright (C) 2019, Xperience AI, all rights reserved.
 Third party copyrights are property of their respective owners.
 
 Redistribution and use in source and binary forms, with or without modification,

SECURITY.md

@@ -0,0 +1,74 @@
+# Security Policy
+
+## Reporting a Vulnerability
+
+If you have information about a security issue or vulnerability in OpenCV, please send an e-mail to security@opencv.org.
+
+Please provide as much information as possible:
+
+- A detailed description of the vulnerability we can use to reproduce your findings.
+
+- Who can exploit this vulnerability and what would they gain. An attack scenario.
+
+- Information about known exploits if any.
+
+A member of the Security Team will review your e-mail and contact you to collaborate on resolving the issue.
+
+## PGP Key
+
+If a security vulnerability report has extremely sensitive information you may encrypt it using our PGP public key:
+
+```
+-----BEGIN PGP PUBLIC KEY BLOCK-----
+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+=Mdna
+-----END PGP PUBLIC KEY BLOCK-----
+```

apps/createsamples/utility.cpp

@@ -895,7 +895,7 @@ void icvGetNextFromBackgroundData( CvBackgroundData* data,
  * #pragma omp parallel
  * {
  *     ...
- *     icvGetBackgourndImage( cvbgdata, cvbgreader, img );
+ *     icvGetBackgroundImage( cvbgdata, cvbgreader, img );
  *     ...
  * }
  * ...
@@ -990,7 +990,7 @@ static int icvInitBackgroundReaders( const char* filename, Size winsize )
 /*
  * icvDestroyBackgroundReaders
  *
- * Finish backgournd reading process
+ * Finish background reading process
  */
 static
 void icvDestroyBackgroundReaders()

cmake/FindCUDA/run_nvcc.cmake

@@ -136,7 +136,7 @@ macro(cuda_execute_process status command)
     # copy and paste a runnable command line.
     set(cuda_execute_process_string)
     foreach(arg ${ARGN})
-      # If there are quotes, excape them, so they come through.
+      # If there are quotes, escape them, so they come through.
       string(REPLACE "\"" "\\\"" arg ${arg})
       # Args with spaces need quotes around them to get them to be parsed as a single argument.
       if(arg MATCHES " ")

cmake/OpenCVCRTLinkage.cmake

@@ -4,16 +4,6 @@ if(NOT MSVC)
   message(FATAL_ERROR "CRT options are available only for MSVC")
 endif()
 
-if (WINRT)
-  if (CMAKE_SYSTEM_VERSION MATCHES 10)
-    add_definitions(/DWINVER=_WIN32_WINNT_WIN10 /DNTDDI_VERSION=NTDDI_WIN10 /D_WIN32_WINNT=_WIN32_WINNT_WIN10)
-  elseif(CMAKE_SYSTEM_VERSION MATCHES 8.1)
-    add_definitions(/DWINVER=_WIN32_WINNT_WINBLUE /DNTDDI_VERSION=NTDDI_WINBLUE /D_WIN32_WINNT=_WIN32_WINNT_WINBLUE)
-  else()
-    add_definitions(/DWINVER=_WIN32_WINNT_WIN8 /DNTDDI_VERSION=NTDDI_WIN8 /D_WIN32_WINNT=_WIN32_WINNT_WIN8)
-  endif()
-endif()
-
 # Removing LNK4075 warnings for debug WinRT builds
 # "LNK4075: ignoring '/INCREMENTAL' due to '/OPT:ICF' specification"
 # "LNK4075: ignoring '/INCREMENTAL' due to '/OPT:REF' specification"

cmake/OpenCVCompilerOptimizations.cmake

@@ -5,13 +5,15 @@
 # AVX / AVX2 / AVX_512F
 # FMA3
 #
+# AVX512 details: https://en.wikipedia.org/wiki/AVX-512#CPUs_with_AVX-512
+#
 # CPU features groups:
 # AVX512_COMMON (Common instructions AVX-512F/CD for all CPUs that support AVX-512)
 # AVX512_KNL (Knights Landing with AVX-512F/CD/ER/PF)
 # AVX512_KNM (Knights Mill with AVX-512F/CD/ER/PF/4FMAPS/4VNNIW/VPOPCNTDQ)
 # AVX512_SKX (Skylake-X with AVX-512F/CD/BW/DQ/VL)
 # AVX512_CNL (Cannon Lake with AVX-512F/CD/BW/DQ/VL/IFMA/VBMI)
-# AVX512_CEL (Cascade Lake with AVX-512F/CD/BW/DQ/VL/IFMA/VBMI/VNNI)
+# AVX512_CLX (Cascade Lake with AVX-512F/CD/BW/DQ/VL/VNNI)
 # AVX512_ICL (Ice Lake with AVX-512F/CD/BW/DQ/VL/IFMA/VBMI/VNNI/VBMI2/BITALG/VPOPCNTDQ/VPCLMULQDQ*/GFNI*/VAES*)
 
 # ppc64le arch:
@@ -43,8 +45,9 @@
 # CPU_{opt}_ENABLED_DEFAULT=ON/OFF - has compiler support without additional flag (CPU_BASELINE_DETECT=ON only)
 
 set(CPU_ALL_OPTIMIZATIONS "SSE;SSE2;SSE3;SSSE3;SSE4_1;SSE4_2;POPCNT;AVX;FP16;AVX2;FMA3;AVX_512F")
-list(APPEND CPU_ALL_OPTIMIZATIONS "AVX512_COMMON;AVX512_KNL;AVX512_KNM;AVX512_SKX;AVX512_CNL;AVX512_CEL;AVX512_ICL")
+list(APPEND CPU_ALL_OPTIMIZATIONS "AVX512_COMMON;AVX512_KNL;AVX512_KNM;AVX512_SKX;AVX512_CNL;AVX512_CLX;AVX512_ICL")
 list(APPEND CPU_ALL_OPTIMIZATIONS NEON VFPV3 FP16)
+list(APPEND CPU_ALL_OPTIMIZATIONS MSA)
 list(APPEND CPU_ALL_OPTIMIZATIONS VSX VSX3)
 list(REMOVE_DUPLICATES CPU_ALL_OPTIMIZATIONS)
 
@@ -162,15 +165,15 @@ elseif(" ${CMAKE_CXX_FLAGS} " MATCHES " -march=native | -xHost | /QxHost ")
 endif()
 
 if(X86 OR X86_64)
-  ocv_update(CPU_KNOWN_OPTIMIZATIONS "SSE;SSE2;SSE3;SSSE3;SSE4_1;POPCNT;SSE4_2;FP16;FMA3;AVX;AVX2;AVX_512F;AVX512_COMMON;AVX512_KNL;AVX512_KNM;AVX512_SKX;AVX512_CNL;AVX512_CEL;AVX512_ICL")
+  ocv_update(CPU_KNOWN_OPTIMIZATIONS "SSE;SSE2;SSE3;SSSE3;SSE4_1;POPCNT;SSE4_2;FP16;FMA3;AVX;AVX2;AVX_512F;AVX512_COMMON;AVX512_KNL;AVX512_KNM;AVX512_SKX;AVX512_CNL;AVX512_CLX;AVX512_ICL")
 
   ocv_update(CPU_AVX512_COMMON_GROUP "AVX_512F;AVX_512CD")
   ocv_update(CPU_AVX512_KNL_GROUP "AVX512_COMMON;AVX512_KNL_EXTRA")
   ocv_update(CPU_AVX512_KNM_GROUP "AVX512_KNL;AVX512_KNM_EXTRA;AVX_512VPOPCNTDQ")
   ocv_update(CPU_AVX512_SKX_GROUP "AVX512_COMMON;AVX_512VL;AVX_512BW;AVX_512DQ")
   ocv_update(CPU_AVX512_CNL_GROUP "AVX512_SKX;AVX_512IFMA;AVX_512VBMI")
-  ocv_update(CPU_AVX512_CEL_GROUP "AVX512_CNL;AVX_512VNNI")
-  ocv_update(CPU_AVX512_ICL_GROUP "AVX512_CEL;AVX_512VBMI2;AVX_512BITALG;AVX_512VPOPCNTDQ") # ? VPCLMULQDQ, GFNI, VAES
+  ocv_update(CPU_AVX512_CLX_GROUP "AVX512_SKX;AVX_512VNNI")
+  ocv_update(CPU_AVX512_ICL_GROUP "AVX512_SKX;AVX_512IFMA;AVX_512VBMI;AVX_512VNNI;AVX_512VBMI2;AVX_512BITALG;AVX_512VPOPCNTDQ") # ? VPCLMULQDQ, GFNI, VAES
 
   ocv_update(CPU_SSE_TEST_FILE "${OpenCV_SOURCE_DIR}/cmake/checks/cpu_sse.cpp")
   ocv_update(CPU_SSE2_TEST_FILE "${OpenCV_SOURCE_DIR}/cmake/checks/cpu_sse2.cpp")
@@ -188,12 +191,12 @@ if(X86 OR X86_64)
   ocv_update(CPU_AVX512_KNM_TEST_FILE "${OpenCV_SOURCE_DIR}/cmake/checks/cpu_avx512knm.cpp")
   ocv_update(CPU_AVX512_SKX_TEST_FILE "${OpenCV_SOURCE_DIR}/cmake/checks/cpu_avx512skx.cpp")
   ocv_update(CPU_AVX512_CNL_TEST_FILE "${OpenCV_SOURCE_DIR}/cmake/checks/cpu_avx512cnl.cpp")
-  ocv_update(CPU_AVX512_CEL_TEST_FILE "${OpenCV_SOURCE_DIR}/cmake/checks/cpu_avx512cel.cpp")
+  ocv_update(CPU_AVX512_CLX_TEST_FILE "${OpenCV_SOURCE_DIR}/cmake/checks/cpu_avx512clx.cpp")
   ocv_update(CPU_AVX512_ICL_TEST_FILE "${OpenCV_SOURCE_DIR}/cmake/checks/cpu_avx512icl.cpp")
 
   if(NOT OPENCV_CPU_OPT_IMPLIES_IGNORE)
-    ocv_update(CPU_AVX512_ICL_IMPLIES "AVX512_CEL")
-    ocv_update(CPU_AVX512_CEL_IMPLIES "AVX512_CNL")
+    ocv_update(CPU_AVX512_ICL_IMPLIES "AVX512_SKX")
+    ocv_update(CPU_AVX512_CLX_IMPLIES "AVX512_SKX")
     ocv_update(CPU_AVX512_CNL_IMPLIES "AVX512_SKX")
     ocv_update(CPU_AVX512_SKX_IMPLIES "AVX512_COMMON")
     ocv_update(CPU_AVX512_KNM_IMPLIES "AVX512_KNL")
@@ -250,7 +253,7 @@ if(X86 OR X86_64)
     ocv_intel_compiler_optimization_option(AVX512_KNM "-xKNM" "/Qx:KNM")
     ocv_intel_compiler_optimization_option(AVX512_SKX "-xSKYLAKE-AVX512" "/Qx:SKYLAKE-AVX512")
     ocv_intel_compiler_optimization_option(AVX512_CNL "-xCANNONLAKE" "/Qx:CANNONLAKE")
-    ocv_intel_compiler_optimization_option(AVX512_CEL "-xCASCADELAKE" "/Qx:CASCADELAKE")
+    ocv_intel_compiler_optimization_option(AVX512_CLX "-xCASCADELAKE" "/Qx:CASCADELAKE")
     ocv_intel_compiler_optimization_option(AVX512_ICL "-xICELAKE-CLIENT" "/Qx:ICELAKE-CLIENT")
   elseif(CV_GCC OR CV_CLANG)
     ocv_update(CPU_AVX2_FLAGS_ON "-mavx2")
@@ -285,7 +288,7 @@ if(X86 OR X86_64)
     ocv_update(CPU_AVX2_FLAGS_ON "/arch:AVX2")
     ocv_update(CPU_AVX_FLAGS_ON "/arch:AVX")
     ocv_update(CPU_FP16_FLAGS_ON "/arch:AVX")
-    if(NOT MSVC64)
+    if(NOT X86_64)
       # 64-bit MSVC compiler uses SSE/SSE2 by default
       ocv_update(CPU_SSE_FLAGS_ON "/arch:SSE")
       ocv_update(CPU_SSE_SUPPORTED ON)
@@ -339,6 +342,11 @@ elseif(ARM OR AARCH64)
     ocv_update(CPU_FP16_IMPLIES "NEON")
     set(CPU_BASELINE "NEON;FP16" CACHE STRING "${HELP_CPU_BASELINE}")
   endif()
+elseif(MIPS)
+  ocv_update(CPU_MSA_TEST_FILE "${OpenCV_SOURCE_DIR}/cmake/checks/cpu_msa.cpp")
+  ocv_update(CPU_KNOWN_OPTIMIZATIONS "MSA")
+  ocv_update(CPU_MSA_FLAGS_ON "-mmsa")
+  set(CPU_BASELINE "DETECT" CACHE STRING "${HELP_CPU_BASELINE}")
 elseif(PPC64LE)
   ocv_update(CPU_KNOWN_OPTIMIZATIONS "VSX;VSX3")
   ocv_update(CPU_VSX_TEST_FILE "${OpenCV_SOURCE_DIR}/cmake/checks/cpu_vsx.cpp")
@@ -706,7 +714,10 @@ macro(ocv_compiler_optimization_process_sources SOURCES_VAR_NAME LIBS_VAR_NAME T
   foreach(OPT ${CPU_DISPATCH_FINAL})
     if(__result_${OPT})
 #message("${OPT}: ${__result_${OPT}}")
-      if(CMAKE_GENERATOR MATCHES "^Visual")
+      if(CMAKE_GENERATOR MATCHES "^Visual"
+          OR OPENCV_CMAKE_CPU_OPTIMIZATIONS_FORCE_TARGETS
+      )
+        # MSVS generator is not able to properly order compilation flags:
         # extra flags are added before common flags, so switching between optimizations doesn't work correctly
         # Also CMAKE_CXX_FLAGS doesn't work (it is directory-based, so add_subdirectory is required)
         add_library(${TARGET_BASE_NAME}_${OPT} OBJECT ${__result_${OPT}})
@@ -854,7 +865,7 @@ macro(__ocv_add_dispatched_file filename target_src_var src_directory dst_direct
 
       if(";${CPU_DISPATCH_FINAL};" MATCHES "${OPT}" OR __CPU_DISPATCH_INCLUDE_ALL)
         if(EXISTS "${src_directory}/${filename}.${OPT_LOWER}.cpp")
-          message(STATUS "Using overrided ${OPT} source: ${src_directory}/${filename}.${OPT_LOWER}.cpp")
+          message(STATUS "Using overridden ${OPT} source: ${src_directory}/${filename}.${OPT_LOWER}.cpp")
         else()
           list(APPEND ${target_src_var} "${__file}")
         endif()

cmake/OpenCVCompilerOptions.cmake

@@ -123,6 +123,9 @@ if(CV_GCC OR CV_CLANG)
   add_extra_compiler_option(-Wsign-promo)
   add_extra_compiler_option(-Wuninitialized)
   add_extra_compiler_option(-Winit-self)
+  if(CV_GCC AND (CMAKE_CXX_COMPILER_VERSION VERSION_GREATER 6.0) AND (CMAKE_CXX_COMPILER_VERSION VERSION_LESS 7.0))
+    add_extra_compiler_option(-Wno-psabi)
+  endif()
   if(HAVE_CXX11)
     if(CMAKE_CXX_COMPILER_ID STREQUAL "GNU" AND NOT ENABLE_PRECOMPILED_HEADERS)
       add_extra_compiler_option(-Wsuggest-override)
@@ -282,11 +285,6 @@ if(MSVC)
   endif()
 endif()
 
-# Adding additional using directory for WindowsPhone 8.0 to get Windows.winmd properly
-if(WINRT_PHONE AND WINRT_8_0)
-  set(OPENCV_EXTRA_CXX_FLAGS "${OPENCV_EXTRA_CXX_FLAGS} /AI\$(WindowsSDK_MetadataPath)")
-endif()
-
 include(cmake/OpenCVCompilerOptimizations.cmake)
 if(COMMAND ocv_compiler_optimization_options)
   ocv_compiler_optimization_options()
@@ -387,6 +385,19 @@ if(MSVC)
   add_definitions(-D_VARIADIC_MAX=10)
 endif()
 
+if(CMAKE_SYSTEM_NAME STREQUAL "Windows")
+  get_directory_property(__DIRECTORY_COMPILE_DEFINITIONS COMPILE_DEFINITIONS)
+  if((NOT " ${CMAKE_CXX_FLAGS} ${CMAKE_CXX_FLAGS_RELEASE} ${OPENCV_EXTRA_CXX_FLAGS} ${OPENCV_EXTRA_FLAGS_RELEASE} ${__DIRECTORY_COMPILE_DEFINITIONS}" MATCHES "_WIN32_WINNT"
+      AND NOT OPENCV_CMAKE_SKIP_MACRO_WIN32_WINNT)
+      OR OPENCV_CMAKE_FORCE_MACRO_WIN32_WINNT
+  )
+    # https://docs.microsoft.com/en-us/cpp/porting/modifying-winver-and-win32-winnt
+    # Target Windows 7 API
+    set(OPENCV_CMAKE_MACRO_WIN32_WINNT "0x0601" CACHE STRING "Value of _WIN32_WINNT macro")
+    add_definitions(-D_WIN32_WINNT=${OPENCV_CMAKE_MACRO_WIN32_WINNT})
+  endif()
+endif()
+
 # Enable compiler options for OpenCV modules/apps/samples only (ignore 3rdparty)
 macro(ocv_add_modules_compiler_options)
   if(MSVC AND NOT OPENCV_SKIP_MSVC_W4_OPTION)

cmake/OpenCVDetectApacheAnt.cmake

@@ -27,7 +27,7 @@ if(ANT_EXECUTABLE)
     unset(ANT_EXECUTABLE CACHE)
   else()
     string(REGEX MATCH "[0-9]+.[0-9]+.[0-9]+" ANT_VERSION "${ANT_VERSION_FULL}")
-    set(ANT_VERSION "${ANT_VERSION}" CACHE INTERNAL "Detected ant vesion")
+    set(ANT_VERSION "${ANT_VERSION}" CACHE INTERNAL "Detected ant version")
 
     message(STATUS "Found apache ant: ${ANT_EXECUTABLE} (${ANT_VERSION})")
   endif()

cmake/OpenCVDetectCUDA.cmake

@@ -12,7 +12,7 @@ endif()
 if(((NOT CMAKE_VERSION VERSION_LESS "3.9.0")  # requires https://gitlab.kitware.com/cmake/cmake/merge_requests/663
       OR OPENCV_CUDA_FORCE_EXTERNAL_CMAKE_MODULE)
     AND NOT OPENCV_CUDA_FORCE_BUILTIN_CMAKE_MODULE)
-  ocv_update(CUDA_LINK_LIBRARIES_KEYWORD "LINK_PRIVATE")
+  ocv_update(CUDA_LINK_LIBRARIES_KEYWORD "PRIVATE")
   find_host_package(CUDA "${MIN_VER_CUDA}" QUIET)
 else()
   # Use OpenCV's patched "FindCUDA" module
@@ -368,7 +368,7 @@ if(HAVE_CUDA)
     endif()
   endif()
 
-  set(OPENCV_LINKER_LIBS ${OPENCV_LINKER_LIBS} ${CUDA_LIBRARIES} ${CUDA_CUDA_LIBRARY} ${CUDA_npp_LIBRARY})
+  set(OPENCV_LINKER_LIBS ${OPENCV_LINKER_LIBS} ${CUDA_LIBRARIES} ${CUDA_npp_LIBRARY})
   if(HAVE_CUBLAS)
     set(OPENCV_LINKER_LIBS ${OPENCV_LINKER_LIBS} ${CUDA_cublas_LIBRARY})
   endif()

cmake/OpenCVDetectCXXCompiler.cmake

@@ -3,15 +3,14 @@
 # - CV_CLANG - Clang-compatible compiler (CMAKE_CXX_COMPILER_ID MATCHES "Clang" - Clang or AppleClang, see CMP0025)
 # - CV_ICC - Intel compiler
 # - MSVC - Microsoft Visual Compiler (CMake variable)
-# - MSVC64 - additional flag, 64-bit
 # - MINGW / CYGWIN / CMAKE_COMPILER_IS_MINGW / CMAKE_COMPILER_IS_CYGWIN (CMake original variables)
-# - MINGW64 - 64-bit
 #
 # CPU Platforms:
 # - X86 / X86_64
 # - ARM - ARM CPU, not defined for AArch64
 # - AARCH64 - ARMv8+ (64-bit)
 # - PPC64 / PPC64LE - PowerPC
+# - MIPS
 #
 # OS:
 # - WIN32 - Windows | MINGW
@@ -21,9 +20,8 @@
 # - APPLE - MacOSX | iOS
 # ----------------------------------------------------------------------------
 
-if(CMAKE_CL_64)
-    set(MSVC64 1)
-endif()
+ocv_declare_removed_variables(MINGW64 MSVC64)
+# do not use (CMake variables): CMAKE_CL_64
 
 if(NOT DEFINED CV_GCC AND CMAKE_CXX_COMPILER_ID MATCHES "GNU")
   set(CV_GCC 1)
@@ -51,7 +49,7 @@ variable_watch(CMAKE_COMPILER_IS_CLANGCC access_CMAKE_COMPILER_IS_CLANGCXX)
 # Detect Intel ICC compiler
 # ----------------------------------------------------------------------------
 if(UNIX)
-  if  (__ICL)
+  if(__ICL)
     set(CV_ICC   __ICL)
   elseif(__ICC)
     set(CV_ICC   __ICC)
@@ -70,51 +68,65 @@ if(MSVC AND CMAKE_C_COMPILER MATCHES "icc|icl")
   set(CV_ICC   __INTEL_COMPILER_FOR_WINDOWS)
 endif()
 
-if(NOT DEFINED CMAKE_CXX_COMPILER_VERSION)
-  message(WARNING "Compiler version is not available: CMAKE_CXX_COMPILER_VERSION is not set")
+if(NOT DEFINED CMAKE_CXX_COMPILER_VERSION
+    AND NOT OPENCV_SUPPRESS_MESSAGE_MISSING_COMPILER_VERSION)
+  message(WARNING "OpenCV: Compiler version is not available: CMAKE_CXX_COMPILER_VERSION is not set")
 endif()
-
-if(WIN32 AND CV_GCC)
-  execute_process(COMMAND ${CMAKE_CXX_COMPILER} -dumpmachine
-                  OUTPUT_VARIABLE OPENCV_GCC_TARGET_MACHINE
-                  OUTPUT_STRIP_TRAILING_WHITESPACE)
-  if(OPENCV_GCC_TARGET_MACHINE MATCHES "amd64|x86_64|AMD64")
-    set(MINGW64 1)
-  endif()
+if((NOT DEFINED CMAKE_SYSTEM_PROCESSOR OR CMAKE_SYSTEM_PROCESSOR STREQUAL "")
+    AND NOT OPENCV_SUPPRESS_MESSAGE_MISSING_CMAKE_SYSTEM_PROCESSOR)
+  message(WARNING "OpenCV: CMAKE_SYSTEM_PROCESSOR is not defined. Perhaps CMake toolchain is broken")
+endif()
+if(NOT DEFINED CMAKE_SIZEOF_VOID_P
+    AND NOT OPENCV_SUPPRESS_MESSAGE_MISSING_CMAKE_SIZEOF_VOID_P)
+  message(WARNING "OpenCV: CMAKE_SIZEOF_VOID_P is not defined. Perhaps CMake toolchain is broken")
 endif()
 
 message(STATUS "Detected processor: ${CMAKE_SYSTEM_PROCESSOR}")
-if(MSVC64 OR MINGW64)
-  set(X86_64 1)
-elseif(MINGW OR (MSVC AND NOT CMAKE_CROSSCOMPILING))
-  set(X86 1)
+if(OPENCV_SKIP_SYSTEM_PROCESSOR_DETECTION)
+  # custom setup: required variables are passed through cache / CMake's command-line
 elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "amd64.*|x86_64.*|AMD64.*")
   set(X86_64 1)
-elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "i686.*|i386.*|x86.*|amd64.*|AMD64.*")
+elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "i686.*|i386.*|x86.*")
   set(X86 1)
+elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "^(aarch64.*|AARCH64.*|arm64.*|ARM64.*)")
+  set(AARCH64 1)
 elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "^(arm.*|ARM.*)")
   set(ARM 1)
-elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "^(aarch64.*|AARCH64.*)")
-  set(AARCH64 1)
 elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "^(powerpc|ppc)64le")
   set(PPC64LE 1)
 elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "^(powerpc|ppc)64")
   set(PPC64 1)
+elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "^(mips.*|MIPS.*)")
+  set(MIPS 1)
+else()
+  if(NOT OPENCV_SUPPRESS_MESSAGE_UNRECOGNIZED_SYSTEM_PROCESSOR)
+    message(WARNING "OpenCV: unrecognized target processor configuration")
+  endif()
 endif()
 
-# Workaround for 32-bit operating systems on x86_64/aarch64 processor
-if(CMAKE_SIZEOF_VOID_P EQUAL 4 AND NOT FORCE_X86_64)
+# Workaround for 32-bit operating systems on x86_64
+if(CMAKE_SIZEOF_VOID_P EQUAL 4 AND X86_64
+    AND NOT FORCE_X86_64  # deprecated (2019-12)
+    AND NOT OPENCV_FORCE_X86_64
+)
   message(STATUS "sizeof(void) = 4 on 64 bit processor. Assume 32-bit compilation mode")
-  if (X86_64)
+  if(X86_64)
     unset(X86_64)
     set(X86 1)
   endif()
-  if (AARCH64)
+endif()
+# Workaround for 32-bit operating systems on aarch64 processor
+if(CMAKE_SIZEOF_VOID_P EQUAL 4 AND AARCH64
+    AND NOT OPENCV_FORCE_AARCH64
+)
+  message(STATUS "sizeof(void) = 4 on 64 bit processor. Assume 32-bit compilation mode")
+  if(AARCH64)
     unset(AARCH64)
     set(ARM 1)
   endif()
 endif()
 
+
 # Similar code exists in OpenCVConfig.cmake
 if(NOT DEFINED OpenCV_STATIC)
   # look for global setting
@@ -128,14 +140,19 @@ endif()
 if(DEFINED OpenCV_ARCH AND DEFINED OpenCV_RUNTIME)
   # custom overridden values
 elseif(MSVC)
-  if(CMAKE_CL_64)
-    set(OpenCV_ARCH x64)
-  elseif((CMAKE_GENERATOR MATCHES "ARM") OR ("${arch_hint}" STREQUAL "ARM") OR (CMAKE_VS_EFFECTIVE_PLATFORMS MATCHES "ARM|arm"))
-    # see Modules/CmakeGenericSystem.cmake
-    set(OpenCV_ARCH ARM)
+  # see Modules/CMakeGenericSystem.cmake
+  if("${CMAKE_GENERATOR}" MATCHES "(Win64|IA64)")
+    set(OpenCV_ARCH "x64")
+  elseif("${CMAKE_GENERATOR_PLATFORM}" MATCHES "ARM64")
+    set(OpenCV_ARCH "ARM64")
+  elseif("${CMAKE_GENERATOR}" MATCHES "ARM")
+    set(OpenCV_ARCH "ARM")
+  elseif("${CMAKE_SIZEOF_VOID_P}" STREQUAL "8")
+    set(OpenCV_ARCH "x64")
   else()
     set(OpenCV_ARCH x86)
   endif()
+
   if(MSVC_VERSION EQUAL 1400)
     set(OpenCV_RUNTIME vc8)
   elseif(MSVC_VERSION EQUAL 1500)
@@ -158,7 +175,7 @@ elseif(MSVC)
 elseif(MINGW)
   set(OpenCV_RUNTIME mingw)
 
-  if(MINGW64)
+  if(CMAKE_SYSTEM_PROCESSOR MATCHES "amd64.*|x86_64.*|AMD64.*")
     set(OpenCV_ARCH x64)
   else()
     set(OpenCV_ARCH x86)
@@ -191,6 +208,25 @@ if(NOT HAVE_CXX11)
     endif()
   endif()
 endif()
+
 if(NOT HAVE_CXX11)
   message(FATAL_ERROR "OpenCV 4.x requires C++11")
 endif()
+
+if((HAVE_CXX11
+        AND NOT MSVC
+        AND NOT (X86 OR X86_64)
+    AND NOT OPENCV_SKIP_LIBATOMIC_COMPILER_CHECK)
+    OR OPENCV_FORCE_LIBATOMIC_COMPILER_CHECK
+)
+  ocv_check_compiler_flag(CXX "" HAVE_CXX_ATOMICS_WITHOUT_LIB "${OpenCV_SOURCE_DIR}/cmake/checks/atomic_check.cpp")
+  if(NOT HAVE_CXX_ATOMICS_WITHOUT_LIB)
+    list(APPEND CMAKE_REQUIRED_LIBRARIES atomic)
+    ocv_check_compiler_flag(CXX "" HAVE_CXX_ATOMICS_WITH_LIB "${OpenCV_SOURCE_DIR}/cmake/checks/atomic_check.cpp")
+    if(HAVE_CXX_ATOMICS_WITH_LIB)
+      list(APPEND OPENCV_LINKER_LIBS atomic)
+    else()
+      message(FATAL_ERROR "C++11 compiler must support std::atomic")
+    endif()
+  endif()
+endif()

cmake/OpenCVDetectInferenceEngine.cmake

@@ -5,7 +5,7 @@
 #
 # Detect parameters:
 # 1. Native cmake IE package:
-#    - enironment variable InferenceEngine_DIR is set to location of cmake module
+#    - environment variable InferenceEngine_DIR is set to location of cmake module
 # 2. Custom location:
 #    - INF_ENGINE_INCLUDE_DIRS - headers search location
 #    - INF_ENGINE_LIB_DIRS     - library search location
@@ -28,6 +28,15 @@ function(add_custom_ie_build _inc _lib _lib_rel _lib_dbg _msg)
     IMPORTED_IMPLIB_DEBUG "${_lib_dbg}"
     INTERFACE_INCLUDE_DIRECTORIES "${_inc}"
   )
+
+  find_library(ie_builder_custom_lib "inference_engine_nn_builder" PATHS "${INF_ENGINE_LIB_DIRS}" NO_DEFAULT_PATH)
+  if(EXISTS "${ie_builder_custom_lib}")
+    add_library(inference_engine_nn_builder UNKNOWN IMPORTED)
+    set_target_properties(inference_engine_nn_builder PROPERTIES
+      IMPORTED_LOCATION "${ie_builder_custom_lib}"
+    )
+  endif()
+
   if(NOT INF_ENGINE_RELEASE VERSION_GREATER "2018050000")
     find_library(INF_ENGINE_OMP_LIBRARY iomp5 PATHS "${INF_ENGINE_OMP_DIR}" NO_DEFAULT_PATH)
     if(NOT INF_ENGINE_OMP_LIBRARY)
@@ -37,7 +46,12 @@ function(add_custom_ie_build _inc _lib _lib_rel _lib_dbg _msg)
     endif()
   endif()
   set(INF_ENGINE_VERSION "Unknown" CACHE STRING "")
-  set(INF_ENGINE_TARGET inference_engine PARENT_SCOPE)
+  set(INF_ENGINE_TARGET inference_engine)
+  if(TARGET inference_engine_nn_builder)
+    list(APPEND INF_ENGINE_TARGET inference_engine_nn_builder)
+    set(_msg "${_msg}, with IE NN Builder API")
+  endif()
+  set(INF_ENGINE_TARGET "${INF_ENGINE_TARGET}" PARENT_SCOPE)
   message(STATUS "Detected InferenceEngine: ${_msg}")
 endfunction()
 
@@ -45,9 +59,9 @@ endfunction()
 
 find_package(InferenceEngine QUIET)
 if(InferenceEngine_FOUND)
-  set(INF_ENGINE_TARGET IE::inference_engine)
+  set(INF_ENGINE_TARGET ${InferenceEngine_LIBRARIES})
   set(INF_ENGINE_VERSION "${InferenceEngine_VERSION}" CACHE STRING "")
-  message(STATUS "Detected InferenceEngine: cmake package")
+  message(STATUS "Detected InferenceEngine: cmake package (${InferenceEngine_VERSION})")
 endif()
 
 if(NOT INF_ENGINE_TARGET AND INF_ENGINE_LIB_DIRS AND INF_ENGINE_INCLUDE_DIRS)
@@ -80,10 +94,22 @@ endif()
 
 if(INF_ENGINE_TARGET)
   if(NOT INF_ENGINE_RELEASE)
-    message(WARNING "InferenceEngine version have not been set, 2019R2 will be used by default. Set INF_ENGINE_RELEASE variable if you experience build errors.")
+    message(WARNING "InferenceEngine version have not been set, 2019R3 will be used by default. Set INF_ENGINE_RELEASE variable if you experience build errors.")
   endif()
-  set(INF_ENGINE_RELEASE "2019020000" CACHE STRING "Force IE version, should be in form YYYYAABBCC (e.g. 2018R2.0.2 -> 2018020002)")
+  set(INF_ENGINE_RELEASE "2019030000" CACHE STRING "Force IE version, should be in form YYYYAABBCC (e.g. 2018R2.0.2 -> 2018020002)")
   set_target_properties(${INF_ENGINE_TARGET} PROPERTIES
     INTERFACE_COMPILE_DEFINITIONS "HAVE_INF_ENGINE=1;INF_ENGINE_RELEASE=${INF_ENGINE_RELEASE}"
   )
 endif()
+
+if(WITH_NGRAPH)
+  find_package(ngraph QUIET)
+  if(ngraph_FOUND)
+    ocv_assert(TARGET ngraph::ngraph)
+    if(INF_ENGINE_RELEASE VERSION_LESS "2019039999")
+      message(WARNING "nGraph is not tested with current InferenceEngine version: INF_ENGINE_RELEASE=${INF_ENGINE_RELEASE}")
+    endif()
+    message(STATUS "Detected ngraph: cmake package (${ngraph_VERSION})")
+    set(HAVE_NGRAPH ON)
+  endif()
+endif()

cmake/OpenCVDetectPython.cmake

@@ -186,11 +186,7 @@ if(NOT ${found})
         else() #debian based assumed, install to the dist-packages.
           set(_packages_path "python${_version_major_minor}/dist-packages")
         endif()
-        if(EXISTS "${CMAKE_INSTALL_PREFIX}/lib${LIB_SUFFIX}/${${packages_path}}")
-          set(_packages_path "lib${LIB_SUFFIX}/${_packages_path}")
-        else()
-          set(_packages_path "lib/${_packages_path}")
-        endif()
+        set(_packages_path "lib/${_packages_path}")
       elseif(CMAKE_HOST_WIN32)
         get_filename_component(_path "${_executable}" PATH)
         file(TO_CMAKE_PATH "${_path}" _path)
@@ -249,7 +245,7 @@ if(NOT ${found})
 
   # Export return values
   set(${found} "${_found}" CACHE INTERNAL "")
-  set(${executable} "${_executable}" CACHE FILEPATH "Path to Python interpretor")
+  set(${executable} "${_executable}" CACHE FILEPATH "Path to Python interpreter")
   set(${version_string} "${_version_string}" CACHE INTERNAL "")
   set(${version_major} "${_version_major}" CACHE INTERNAL "")
   set(${version_minor} "${_version_minor}" CACHE INTERNAL "")

cmake/OpenCVDownload.cmake

@@ -22,6 +22,9 @@ set(OPENCV_DOWNLOAD_PATH "${OpenCV_SOURCE_DIR}/.cache" CACHE PATH "${HELP_OPENCV
 set(OPENCV_DOWNLOAD_LOG "${OpenCV_BINARY_DIR}/CMakeDownloadLog.txt")
 set(OPENCV_DOWNLOAD_WITH_CURL "${OpenCV_BINARY_DIR}/download_with_curl.sh")
 set(OPENCV_DOWNLOAD_WITH_WGET "${OpenCV_BINARY_DIR}/download_with_wget.sh")
+set(OPENCV_DOWNLOAD_TRIES_LIST 1 CACHE STRING "List of download tries") # a list
+set(OPENCV_DOWNLOAD_PARAMS INACTIVITY_TIMEOUT 60 TIMEOUT 600 CACHE STRING "Download parameters to be passed to file(DOWNLAOD ...)")
+mark_as_advanced(OPENCV_DOWNLOAD_TRIES_LIST OPENCV_DOWNLOAD_PARAMS)
 
 # Init download cache directory and log file and helper scripts
 if(NOT EXISTS "${OPENCV_DOWNLOAD_PATH}")
@@ -154,11 +157,17 @@ function(ocv_download)
   # Download
   if(NOT EXISTS "${CACHE_CANDIDATE}")
     ocv_download_log("#cmake_download \"${CACHE_CANDIDATE}\" \"${DL_URL}\"")
-    file(DOWNLOAD "${DL_URL}" "${CACHE_CANDIDATE}"
-         INACTIVITY_TIMEOUT 60
-         TIMEOUT 600
-         STATUS status
-         LOG __log)
+    foreach(try ${OPENCV_DOWNLOAD_TRIES_LIST})
+      ocv_download_log("#try ${try}")
+      file(DOWNLOAD "${DL_URL}" "${CACHE_CANDIDATE}"
+           STATUS status
+           LOG __log
+           ${OPENCV_DOWNLOAD_PARAMS})
+      if(status EQUAL 0)
+        break()
+      endif()
+      message(STATUS "Try ${try} failed")
+    endforeach()
     if(NOT OPENCV_SKIP_FILE_DOWNLOAD_DUMP)  # workaround problem with old CMake versions: "Invalid escape sequence"
       string(LENGTH "${__log}" __log_length)
       if(__log_length LESS 65536)
@@ -195,8 +204,8 @@ For details please refer to the download log file:
 ${OPENCV_DOWNLOAD_LOG}
 ")
       # write helper scripts for failed downloads
-      file(APPEND "${OPENCV_DOWNLOAD_WITH_CURL}" "curl --output \"${CACHE_CANDIDATE}\" \"${DL_URL}\"\n")
-      file(APPEND "${OPENCV_DOWNLOAD_WITH_WGET}" "wget -O \"${CACHE_CANDIDATE}\" \"${DL_URL}\"\n")
+      file(APPEND "${OPENCV_DOWNLOAD_WITH_CURL}" "curl --create-dirs --output \"${CACHE_CANDIDATE}\" \"${DL_URL}\"\n")
+      file(APPEND "${OPENCV_DOWNLOAD_WITH_WGET}" "mkdir -p $(dirname ${CACHE_CANDIDATE}) && wget -O \"${CACHE_CANDIDATE}\" \"${DL_URL}\"\n")
       return()
     endif()
 

cmake/OpenCVFindIPP.cmake

@@ -40,11 +40,8 @@ if (X86 AND UNIX AND NOT APPLE AND NOT ANDROID AND BUILD_SHARED_LIBS)
 endif()
 
 set(IPP_X64 0)
-if(CMAKE_CXX_SIZEOF_DATA_PTR EQUAL 8)
-    set(IPP_X64 1)
-endif()
-if(CMAKE_CL_64)
-    set(IPP_X64 1)
+if(X86_64)
+  set(IPP_X64 1)
 endif()
 
 # This function detects Intel IPP version by analyzing .h file

cmake/OpenCVFindIPPIW.cmake

@@ -29,6 +29,16 @@ macro(ippiw_debugmsg MESSAGE)
     message(STATUS "${MESSAGE}")
   endif()
 endmacro()
+
+macro(ippiw_done)
+  foreach(__file ${IPP_IW_LICENSE_FILES})
+    if(EXISTS "${__file}")
+      ocv_install_3rdparty_licenses(ippiw "${__file}")
+    endif()
+  endforeach()
+  return()
+endmacro()
+
 file(TO_CMAKE_PATH "${IPPROOT}" IPPROOT)
 
 # This function detects Intel IPP Integration Wrappers version by analyzing .h file
@@ -81,7 +91,7 @@ macro(ippiw_setup PATH BUILD)
           if(EXISTS "${FILE}")
             set(HAVE_IPP_IW_LL 1)
           endif()
-          return()
+          ippiw_done()
         else()
           ippiw_debugmsg("sources\tno")
         endif()
@@ -120,7 +130,7 @@ macro(ippiw_setup PATH BUILD)
           if(EXISTS "${FILE}")
             set(HAVE_IPP_IW_LL 1)
           endif()
-          return()
+          ippiw_done()
         else()
           ippiw_debugmsg("binaries\tno")
         endif()
@@ -147,14 +157,12 @@ if(BUILD_IPP_IW)
   ippiw_setup("${OpenCV_SOURCE_DIR}/3rdparty/ippiw" 1)
 
   set(IPPIW_ROOT "${IPPROOT}/../iw")
-  ocv_install_3rdparty_licenses(ippiw
-    "${IPPIW_ROOT}/../support.txt"
-    "${IPPIW_ROOT}/../third-party-programs.txt")
-  if(WIN32)
-    ocv_install_3rdparty_licenses(ippiw "${IPPIW_ROOT}/../EULA.rtf")
-  else()
-    ocv_install_3rdparty_licenses(ippiw "${IPPIW_ROOT}/../EULA.txt")
-  endif()
+  set(IPP_IW_LICENSE_FILES ${IPP_IW_LICENSE_FILES_EXTRA}
+      "${IPPIW_ROOT}/../support.txt"
+      "${IPPIW_ROOT}/../third-party-programs.txt"
+      "${IPPIW_ROOT}/../EULA.rtf"
+      "${IPPIW_ROOT}/../EULA.txt"
+  )
 
   # Package sources
   get_filename_component(__PATH "${IPPROOT}/../iw/" ABSOLUTE)
@@ -167,10 +175,11 @@ if(BUILD_IPP_IW)
     include("${OpenCV_SOURCE_DIR}/3rdparty/ippicv/ippicv.cmake")
     download_ippicv(TEMP_ROOT)
     set(IPPIW_ROOT "${TEMP_ROOT}/iw/")
-    ocv_install_3rdparty_licenses(ippiw
-      "${IPPIW_ROOT}/../EULA.txt"
-      "${IPPIW_ROOT}/../support.txt"
-      "${IPPIW_ROOT}/../third-party-programs.txt")
+    set(IPP_IW_LICENSE_FILES ${IPP_IW_LICENSE_FILES_EXTRA}
+        "${IPPIW_ROOT}/../EULA.txt"
+        "${IPPIW_ROOT}/../support.txt"
+        "${IPPIW_ROOT}/../third-party-programs.txt"
+    )
 
     ippiw_setup("${IPPIW_ROOT}" 1)
   endif()

cmake/OpenCVFindMKL.cmake

@@ -133,7 +133,7 @@ message(STATUS "Found MKL ${MKL_VERSION_STR} at: ${MKL_ROOT_DIR}")
 set(HAVE_MKL ON)
 set(MKL_ROOT_DIR "${MKL_ROOT_DIR}" CACHE PATH "Path to MKL directory")
 set(MKL_INCLUDE_DIRS "${MKL_INCLUDE_DIRS}" CACHE PATH "Path to MKL include directory")
-set(MKL_LIBRARIES "${MKL_LIBRARIES}" CACHE STRING "MKL libarries")
+set(MKL_LIBRARIES "${MKL_LIBRARIES}" CACHE STRING "MKL libraries")
 if(UNIX AND NOT MKL_LIBRARIES_DONT_HACK)
     #it's ugly but helps to avoid cyclic lib problem
     set(MKL_LIBRARIES ${MKL_LIBRARIES} ${MKL_LIBRARIES} ${MKL_LIBRARIES} "-lpthread" "-lm" "-ldl")

cmake/OpenCVFindOpenEXR.cmake

@@ -15,60 +15,94 @@ file(TO_CMAKE_PATH "$ENV{ProgramFiles}" ProgramFiles_ENV_PATH)
 
 if(WIN32)
     SET(OPENEXR_ROOT "C:/Deploy" CACHE STRING "Path to the OpenEXR \"Deploy\" folder")
-    if(CMAKE_CL_64)
+    if(X86_64)
         SET(OPENEXR_LIBSEARCH_SUFFIXES x64/Release x64 x64/Debug)
     elseif(MSVC)
         SET(OPENEXR_LIBSEARCH_SUFFIXES Win32/Release Win32 Win32/Debug)
     endif()
-else()
-    set(OPENEXR_ROOT "")
 endif()
 
-SET(LIBRARY_PATHS
-    /usr/lib
-    /usr/local/lib
-    /sw/lib
-    /opt/local/lib
-    "${ProgramFiles_ENV_PATH}/OpenEXR/lib/static"
-    "${OPENEXR_ROOT}/lib")
+SET(SEARCH_PATHS
+    "${OPENEXR_ROOT}"
+    /usr
+    /usr/local
+    /sw
+    /opt
+    "${ProgramFiles_ENV_PATH}/OpenEXR")
 
-FIND_PATH(OPENEXR_INCLUDE_PATH ImfRgbaFile.h
-    PATH_SUFFIXES OpenEXR
-    PATHS
-    /usr/include
-    /usr/local/include
-    /sw/include
-    /opt/local/include
-    "${ProgramFiles_ENV_PATH}/OpenEXR/include"
-    "${OPENEXR_ROOT}/include")
+MACRO(FIND_OPENEXR_LIBRARY LIBRARY_NAME LIBRARY_SUFFIX)
+    string(TOUPPER "${LIBRARY_NAME}" LIBRARY_NAME_UPPER)
+    FIND_LIBRARY(OPENEXR_${LIBRARY_NAME_UPPER}_LIBRARY
+        NAMES ${LIBRARY_NAME}${LIBRARY_SUFFIX}
+        PATH_SUFFIXES ${OPENEXR_LIBSEARCH_SUFFIXES}
+        NO_DEFAULT_PATH
+        PATHS "${SEARCH_PATH}/lib" "${SEARCH_PATH}/lib/static")
+ENDMACRO()
 
-FIND_LIBRARY(OPENEXR_HALF_LIBRARY
-    NAMES Half
-    PATH_SUFFIXES ${OPENEXR_LIBSEARCH_SUFFIXES}
-    PATHS ${LIBRARY_PATHS})
+FOREACH(SEARCH_PATH ${SEARCH_PATHS})
+    FIND_PATH(OPENEXR_INCLUDE_PATH ImfRgbaFile.h
+        PATH_SUFFIXES OpenEXR
+        NO_DEFAULT_PATH
+        PATHS
+        "${SEARCH_PATH}/include")
 
-FIND_LIBRARY(OPENEXR_IEX_LIBRARY
-    NAMES Iex
-    PATH_SUFFIXES ${OPENEXR_LIBSEARCH_SUFFIXES}
-    PATHS ${LIBRARY_PATHS})
+    IF (OPENEXR_INCLUDE_PATH)
+        SET(OPENEXR_VERSION_FILE "${OPENEXR_INCLUDE_PATH}/OpenEXRConfig.h")
+        IF (EXISTS ${OPENEXR_VERSION_FILE})
+            FILE (STRINGS ${OPENEXR_VERSION_FILE} contents REGEX "#define OPENEXR_VERSION_MAJOR ")
+            IF (${contents} MATCHES "#define OPENEXR_VERSION_MAJOR ([0-9]+)")
+                SET(OPENEXR_VERSION_MAJOR "${CMAKE_MATCH_1}")
+            ENDIF ()
+            FILE (STRINGS ${OPENEXR_VERSION_FILE} contents REGEX "#define OPENEXR_VERSION_MINOR ")
+            IF (${contents} MATCHES "#define OPENEXR_VERSION_MINOR ([0-9]+)")
+                SET(OPENEXR_VERSION_MINOR "${CMAKE_MATCH_1}")
+            ENDIF ()
+        ENDIF ()
+    ENDIF ()
 
-FIND_LIBRARY(OPENEXR_IMATH_LIBRARY
-    NAMES Imath
-    PATH_SUFFIXES ${OPENEXR_LIBSEARCH_SUFFIXES}
-    PATHS ${LIBRARY_PATHS})
+    IF (OPENEXR_VERSION_MAJOR AND OPENEXR_VERSION_MINOR)
+        set(OPENEXR_VERSION "${OPENEXR_VERSION_MAJOR}_${OPENEXR_VERSION_MINOR}")
+    ENDIF ()
 
-FIND_LIBRARY(OPENEXR_ILMIMF_LIBRARY
-    NAMES IlmImf
-    PATH_SUFFIXES ${OPENEXR_LIBSEARCH_SUFFIXES}
-    PATHS ${LIBRARY_PATHS})
+    SET(LIBRARY_SUFFIXES
+        "-${OPENEXR_VERSION}"
+        "-${OPENEXR_VERSION}_s"
+        "-${OPENEXR_VERSION}_d"
+        "-${OPEXEXR_VERSION}_s_d"
+        ""
+        "_s"
+        "_d"
+        "_s_d")
 
-FIND_LIBRARY(OPENEXR_ILMTHREAD_LIBRARY
-    NAMES IlmThread
-    PATH_SUFFIXES ${OPENEXR_LIBSEARCH_SUFFIXES}
-    PATHS ${LIBRARY_PATHS})
+    FOREACH(LIBRARY_SUFFIX ${LIBRARY_SUFFIXES})
+        FIND_OPENEXR_LIBRARY("Half" ${LIBRARY_SUFFIX})
+        FIND_OPENEXR_LIBRARY("Iex" ${LIBRARY_SUFFIX})
+        FIND_OPENEXR_LIBRARY("Imath" ${LIBRARY_SUFFIX})
+        FIND_OPENEXR_LIBRARY("IlmImf" ${LIBRARY_SUFFIX})
+        FIND_OPENEXR_LIBRARY("IlmThread" ${LIBRARY_SUFFIX})
+        IF (OPENEXR_INCLUDE_PATH AND OPENEXR_IMATH_LIBRARY AND OPENEXR_ILMIMF_LIBRARY AND OPENEXR_IEX_LIBRARY AND OPENEXR_HALF_LIBRARY)
+            SET(OPENEXR_FOUND TRUE)
+            BREAK()
+        ENDIF()
+        UNSET(OPENEXR_IMATH_LIBRARY)
+        UNSET(OPENEXR_ILMIMF_LIBRARY)
+        UNSET(OPENEXR_IEX_LIBRARY)
+        UNSET(OPENEXR_ILMTHREAD_LIBRARY)
+        UNSET(OPENEXR_HALF_LIBRARY)
+    ENDFOREACH()
 
-IF (OPENEXR_INCLUDE_PATH AND OPENEXR_IMATH_LIBRARY AND OPENEXR_ILMIMF_LIBRARY AND OPENEXR_IEX_LIBRARY AND OPENEXR_HALF_LIBRARY)
-    SET(OPENEXR_FOUND TRUE)
+    IF (OPENEXR_FOUND)
+        BREAK()
+    ENDIF()
+
+    UNSET(OPENEXR_INCLUDE_PATH)
+    UNSET(OPENEXR_VERSION_FILE)
+    UNSET(OPENEXR_VERSION_MAJOR)
+    UNSET(OPENEXR_VERSION_MINOR)
+    UNSET(OPENEXR_VERSION)
+ENDFOREACH()
+
+IF (OPENEXR_FOUND)
     SET(OPENEXR_INCLUDE_PATHS ${OPENEXR_INCLUDE_PATH} CACHE PATH "The include paths needed to use OpenEXR")
     SET(OPENEXR_LIBRARIES ${OPENEXR_IMATH_LIBRARY} ${OPENEXR_ILMIMF_LIBRARY} ${OPENEXR_IEX_LIBRARY} ${OPENEXR_HALF_LIBRARY} ${OPENEXR_ILMTHREAD_LIBRARY} CACHE STRING "The libraries needed to use OpenEXR" FORCE)
 ENDIF ()

cmake/OpenCVGenConfig.cmake

@@ -122,13 +122,10 @@ endif()
 #  Part 3/3: ${BIN_DIR}/win-install/OpenCVConfig.cmake  -> For use within binary installers/packages
 # --------------------------------------------------------------------------------------------
 if(WIN32)
-  if(CMAKE_HOST_SYSTEM_NAME MATCHES Windows)
-    if(BUILD_SHARED_LIBS)
-      set(_lib_suffix "lib")
-    else()
-      set(_lib_suffix "staticlib")
-    endif()
-    ocv_gen_config("${CMAKE_BINARY_DIR}/win-install" "${OpenCV_INSTALL_BINARIES_PREFIX}${_lib_suffix}" "OpenCVConfig.root-WIN32.cmake.in")
+  if(CMAKE_HOST_SYSTEM_NAME MATCHES Windows AND NOT OPENCV_SKIP_CMAKE_ROOT_CONFIG)
+    ocv_gen_config("${CMAKE_BINARY_DIR}/win-install"
+                   "${OPENCV_INSTALL_BINARIES_PREFIX}${OPENCV_INSTALL_BINARIES_SUFFIX}"
+                   "OpenCVConfig.root-WIN32.cmake.in")
   else()
     ocv_gen_config("${CMAKE_BINARY_DIR}/win-install" "" "")
   endif()

cmake/OpenCVInstallLayout.cmake

@@ -0,0 +1,101 @@
+# message(STATUS "Initial install layout:")
+# ocv_cmake_dump_vars("OPENCV_.*_INSTALL_PATH")
+
+if(ANDROID)
+
+  ocv_update(OPENCV_BIN_INSTALL_PATH            "sdk/native/bin/${ANDROID_NDK_ABI_NAME}")
+  ocv_update(OPENCV_TEST_INSTALL_PATH           "${OPENCV_BIN_INSTALL_PATH}")
+  ocv_update(OPENCV_SAMPLES_BIN_INSTALL_PATH    "sdk/native/samples/${ANDROID_NDK_ABI_NAME}")
+  ocv_update(OPENCV_LIB_INSTALL_PATH            "sdk/native/libs/${ANDROID_NDK_ABI_NAME}")
+  ocv_update(OPENCV_LIB_ARCHIVE_INSTALL_PATH    "sdk/native/staticlibs/${ANDROID_NDK_ABI_NAME}")
+  ocv_update(OPENCV_3P_LIB_INSTALL_PATH         "sdk/native/3rdparty/libs/${ANDROID_NDK_ABI_NAME}")
+  ocv_update(OPENCV_CONFIG_INSTALL_PATH         "sdk/native/jni")
+  ocv_update(OPENCV_INCLUDE_INSTALL_PATH        "sdk/native/jni/include")
+  ocv_update(OPENCV_OTHER_INSTALL_PATH          "sdk/etc")
+  ocv_update(OPENCV_SAMPLES_SRC_INSTALL_PATH    "samples/native")
+  ocv_update(OPENCV_LICENSES_INSTALL_PATH       "${OPENCV_OTHER_INSTALL_PATH}/licenses")
+  ocv_update(OPENCV_TEST_DATA_INSTALL_PATH      "${OPENCV_OTHER_INSTALL_PATH}/testdata")
+  ocv_update(OPENCV_DOC_INSTALL_PATH            "doc")
+  ocv_update(OPENCV_JAR_INSTALL_PATH            ".")
+  ocv_update(OPENCV_JNI_INSTALL_PATH            "${OPENCV_LIB_INSTALL_PATH}")
+  ocv_update(OPENCV_JNI_BIN_INSTALL_PATH        "${OPENCV_JNI_INSTALL_PATH}")
+
+elseif(WIN32 AND CMAKE_HOST_SYSTEM_NAME MATCHES Windows)
+
+  if(DEFINED OpenCV_RUNTIME AND DEFINED OpenCV_ARCH)
+    ocv_update(OPENCV_INSTALL_BINARIES_PREFIX "${OpenCV_ARCH}/${OpenCV_RUNTIME}/")
+  else()
+    message(STATUS "Can't detect runtime and/or arch")
+    ocv_update(OPENCV_INSTALL_BINARIES_PREFIX "")
+  endif()
+  if(OpenCV_STATIC)
+    ocv_update(OPENCV_INSTALL_BINARIES_SUFFIX "staticlib")
+  else()
+    ocv_update(OPENCV_INSTALL_BINARIES_SUFFIX "lib")
+  endif()
+  if(INSTALL_CREATE_DISTRIB)
+    set(_jni_suffix "/${OpenCV_ARCH}")
+  else()
+    set(_jni_suffix "")
+  endif()
+
+  ocv_update(OPENCV_BIN_INSTALL_PATH           "${OPENCV_INSTALL_BINARIES_PREFIX}bin")
+  ocv_update(OPENCV_TEST_INSTALL_PATH          "${OPENCV_BIN_INSTALL_PATH}")
+  ocv_update(OPENCV_SAMPLES_BIN_INSTALL_PATH   "${OPENCV_INSTALL_BINARIES_PREFIX}samples")
+  ocv_update(OPENCV_LIB_INSTALL_PATH           "${OPENCV_INSTALL_BINARIES_PREFIX}${OPENCV_INSTALL_BINARIES_SUFFIX}")
+  ocv_update(OPENCV_LIB_ARCHIVE_INSTALL_PATH   "${OPENCV_LIB_INSTALL_PATH}")
+  ocv_update(OPENCV_3P_LIB_INSTALL_PATH        "${OPENCV_INSTALL_BINARIES_PREFIX}staticlib")
+  ocv_update(OPENCV_CONFIG_INSTALL_PATH        ".")
+  ocv_update(OPENCV_INCLUDE_INSTALL_PATH       "include")
+  ocv_update(OPENCV_OTHER_INSTALL_PATH         "etc")
+  ocv_update(OPENCV_SAMPLES_SRC_INSTALL_PATH   "samples")
+  ocv_update(OPENCV_LICENSES_INSTALL_PATH      "${OPENCV_OTHER_INSTALL_PATH}/licenses")
+  ocv_update(OPENCV_TEST_DATA_INSTALL_PATH     "testdata")
+  ocv_update(OPENCV_DOC_INSTALL_PATH           "doc")
+  ocv_update(OPENCV_JAR_INSTALL_PATH           "java")
+  ocv_update(OPENCV_JNI_INSTALL_PATH           "java${_jni_suffix}")
+  ocv_update(OPENCV_JNI_BIN_INSTALL_PATH       "${OPENCV_JNI_INSTALL_PATH}")
+
+else() # UNIX
+
+  include(GNUInstallDirs)
+  ocv_update(OPENCV_BIN_INSTALL_PATH           "bin")
+  ocv_update(OPENCV_TEST_INSTALL_PATH          "${OPENCV_BIN_INSTALL_PATH}")
+  ocv_update(OPENCV_SAMPLES_BIN_INSTALL_PATH   "${OPENCV_BIN_INSTALL_PATH}")
+  ocv_update(OPENCV_LIB_INSTALL_PATH           "${CMAKE_INSTALL_LIBDIR}")
+  ocv_update(OPENCV_LIB_ARCHIVE_INSTALL_PATH   "${OPENCV_LIB_INSTALL_PATH}")
+  ocv_update(OPENCV_3P_LIB_INSTALL_PATH        "${OPENCV_LIB_INSTALL_PATH}/opencv4/3rdparty")
+  ocv_update(OPENCV_CONFIG_INSTALL_PATH        "${OPENCV_LIB_INSTALL_PATH}/cmake/opencv4")
+  ocv_update(OPENCV_INCLUDE_INSTALL_PATH       "${CMAKE_INSTALL_INCLUDEDIR}/opencv4")
+  ocv_update(OPENCV_OTHER_INSTALL_PATH         "${CMAKE_INSTALL_DATAROOTDIR}/opencv4")
+  ocv_update(OPENCV_SAMPLES_SRC_INSTALL_PATH   "${OPENCV_OTHER_INSTALL_PATH}/samples")
+  ocv_update(OPENCV_LICENSES_INSTALL_PATH      "${CMAKE_INSTALL_DATAROOTDIR}/licenses/opencv4")
+  ocv_update(OPENCV_TEST_DATA_INSTALL_PATH     "${OPENCV_OTHER_INSTALL_PATH}/testdata")
+  ocv_update(OPENCV_DOC_INSTALL_PATH           "${CMAKE_INSTALL_DATAROOTDIR}/doc/opencv4")
+  ocv_update(OPENCV_JAR_INSTALL_PATH           "${CMAKE_INSTALL_DATAROOTDIR}/java/opencv4")
+  ocv_update(OPENCV_JNI_INSTALL_PATH           "${OPENCV_JAR_INSTALL_PATH}")
+  ocv_update(OPENCV_JNI_BIN_INSTALL_PATH       "${OPENCV_JNI_INSTALL_PATH}")
+
+endif()
+
+ocv_update(CMAKE_INSTALL_RPATH "${CMAKE_INSTALL_PREFIX}/${OPENCV_LIB_INSTALL_PATH}")
+set(CMAKE_INSTALL_RPATH_USE_LINK_PATH TRUE)
+
+if(INSTALL_TO_MANGLED_PATHS)
+  foreach(v
+      OPENCV_INCLUDE_INSTALL_PATH
+      # file names include version (.so/.dll): OPENCV_LIB_INSTALL_PATH
+      OPENCV_CONFIG_INSTALL_PATH
+      OPENCV_3P_LIB_INSTALL_PATH
+      OPENCV_SAMPLES_SRC_INSTALL_PATH
+      OPENCV_DOC_INSTALL_PATH
+      # JAR file name includes version: OPENCV_JAR_INSTALL_PATH
+      OPENCV_TEST_DATA_INSTALL_PATH
+      OPENCV_OTHER_INSTALL_PATH
+    )
+    string(REGEX REPLACE "opencv[0-9]*" "opencv-${OPENCV_VERSION}" ${v} "${${v}}")
+  endforeach()
+endif()
+
+# message(STATUS "Final install layout:")
+# ocv_cmake_dump_vars("OPENCV_.*_INSTALL_PATH")

cmake/OpenCVMinDepVersions.cmake

@@ -1,6 +1,8 @@
-set(MIN_VER_CMAKE 3.5.1)
+if(NOT DEFINED MIN_VER_CMAKE)
+  set(MIN_VER_CMAKE 3.5.1)
+endif()
 set(MIN_VER_CUDA 6.5)
-set(MIN_VER_CUDNN 6)
+set(MIN_VER_CUDNN 7.5)
 set(MIN_VER_PYTHON2 2.7)
 set(MIN_VER_PYTHON3 3.2)
 set(MIN_VER_ZLIB 1.2.3)

cmake/OpenCVModule.cmake

@@ -63,7 +63,6 @@ foreach(mod ${OPENCV_MODULES_BUILD} ${OPENCV_MODULES_DISABLED_USER} ${OPENCV_MOD
   unset(OPENCV_MODULE_${mod}_PRIVATE_OPT_DEPS CACHE)
   unset(OPENCV_MODULE_${mod}_LINK_DEPS CACHE)
   unset(OPENCV_MODULE_${mod}_WRAPPERS CACHE)
-  unset(OPENCV_DEPENDANT_TARGETS_${mod} CACHE)
 endforeach()
 
 # clean modules info which needs to be recalculated
@@ -852,7 +851,7 @@ macro(ocv_create_module)
     set(the_module_target ${the_module})
   endif()
 
-  if(WINRT)
+  if(WINRT AND BUILD_TESTS)
     # removing APPCONTAINER from modules to run from console
     # in case of usual starting of WinRT test apps output is missing
     # so starting of console version w/o APPCONTAINER is required to get test results
@@ -937,11 +936,15 @@ macro(_ocv_create_module)
   set_source_files_properties(${OPENCV_MODULE_${the_module}_HEADERS} ${OPENCV_MODULE_${the_module}_SOURCES} ${${the_module}_pch}
     PROPERTIES LABELS "${OPENCV_MODULE_${the_module}_LABEL};Module")
 
-  ocv_target_link_libraries(${the_module} LINK_PUBLIC ${OPENCV_MODULE_${the_module}_DEPS_TO_LINK})
-  ocv_target_link_libraries(${the_module} LINK_PUBLIC ${OPENCV_MODULE_${the_module}_DEPS_EXT})
-  ocv_target_link_libraries(${the_module} LINK_PRIVATE ${OPENCV_LINKER_LIBS} ${OPENCV_HAL_LINKER_LIBS} ${IPP_LIBS} ${ARGN})
+  ocv_target_link_libraries(${the_module} PUBLIC    ${OPENCV_MODULE_${the_module}_DEPS_TO_LINK}
+                                          INTERFACE ${OPENCV_MODULE_${the_module}_DEPS_TO_LINK}
+  )
+  ocv_target_link_libraries(${the_module} PUBLIC    ${OPENCV_MODULE_${the_module}_DEPS_EXT}
+                                          INTERFACE ${OPENCV_MODULE_${the_module}_DEPS_EXT}
+  )
+  ocv_target_link_libraries(${the_module} PRIVATE ${OPENCV_LINKER_LIBS} ${OPENCV_HAL_LINKER_LIBS} ${IPP_LIBS} ${ARGN})
   if (HAVE_CUDA)
-    ocv_target_link_libraries(${the_module} LINK_PRIVATE ${CUDA_LIBRARIES} ${CUDA_npp_LIBRARY})
+    ocv_target_link_libraries(${the_module} PRIVATE ${CUDA_LIBRARIES} ${CUDA_npp_LIBRARY})
   endif()
 
   if(OPENCV_MODULE_${the_module}_COMPILE_DEFINITIONS)
@@ -1150,7 +1153,7 @@ function(ocv_add_perf_tests)
       source_group("Src" FILES "${${the_target}_pch}")
       ocv_add_executable(${the_target} ${OPENCV_PERF_${the_module}_SOURCES} ${${the_target}_pch})
       ocv_target_include_modules(${the_target} ${perf_deps})
-      ocv_target_link_libraries(${the_target} LINK_PRIVATE ${perf_deps} ${OPENCV_MODULE_${the_module}_DEPS} ${OPENCV_LINKER_LIBS} ${OPENCV_PERF_${the_module}_DEPS})
+      ocv_target_link_libraries(${the_target} PRIVATE ${perf_deps} ${OPENCV_MODULE_${the_module}_DEPS} ${OPENCV_LINKER_LIBS} ${OPENCV_PERF_${the_module}_DEPS})
       add_dependencies(opencv_perf_tests ${the_target})
 
       if(TARGET opencv_videoio_plugins)
@@ -1206,10 +1209,6 @@ endfunction()
 function(ocv_add_accuracy_tests)
   ocv_debug_message("ocv_add_accuracy_tests(" ${ARGN} ")")
 
-  if(WINRT)
-    set(OPENCV_DEBUG_POSTFIX "")
-  endif()
-
   set(test_path "${CMAKE_CURRENT_LIST_DIR}/test")
   if(BUILD_TESTS AND EXISTS "${test_path}")
     __ocv_parse_test_sources(TEST ${ARGN})
@@ -1244,7 +1243,7 @@ function(ocv_add_accuracy_tests)
       if(EXISTS "${CMAKE_CURRENT_BINARY_DIR}/test")
         ocv_target_include_directories(${the_target} "${CMAKE_CURRENT_BINARY_DIR}/test")
       endif()
-      ocv_target_link_libraries(${the_target} LINK_PRIVATE ${test_deps} ${OPENCV_MODULE_${the_module}_DEPS} ${OPENCV_LINKER_LIBS} ${OPENCV_TEST_${the_module}_DEPS})
+      ocv_target_link_libraries(${the_target} PRIVATE ${test_deps} ${OPENCV_MODULE_${the_module}_DEPS} ${OPENCV_LINKER_LIBS} ${OPENCV_TEST_${the_module}_DEPS})
       add_dependencies(opencv_tests ${the_target})
 
       if(TARGET opencv_videoio_plugins)
@@ -1314,7 +1313,7 @@ function(ocv_add_samples)
 
         ocv_add_executable(${the_target} "${source}")
         ocv_target_include_modules(${the_target} ${samples_deps})
-        ocv_target_link_libraries(${the_target} LINK_PRIVATE ${samples_deps})
+        ocv_target_link_libraries(${the_target} PRIVATE ${samples_deps})
 
         set_target_properties(${the_target} PROPERTIES
           PROJECT_LABEL "(sample) ${name}"

cmake/OpenCVPCHSupport.cmake

@@ -125,11 +125,11 @@ MACRO(_PCH_GET_COMPILE_COMMAND out_command _input _output)
             STRING(REGEX REPLACE "^ +" "" pchsupport_compiler_cxx_arg1 ${CMAKE_CXX_COMPILER_ARG1})
 
             SET(${out_command}
-              ${CMAKE_CXX_COMPILER} ${pchsupport_compiler_cxx_arg1} ${_compile_FLAGS} -x c++-header -o ${_output} ${_input}
+              ${CMAKE_CXX_COMPILER} ${pchsupport_compiler_cxx_arg1} ${_compile_FLAGS} -x c++-header -o ${_output} -c ${_input}
               )
         ELSE(CMAKE_CXX_COMPILER_ARG1)
             SET(${out_command}
-              ${CMAKE_CXX_COMPILER}  ${_compile_FLAGS} -x c++-header -o ${_output} ${_input}
+              ${CMAKE_CXX_COMPILER}  ${_compile_FLAGS} -x c++-header -o ${_output} -c ${_input}
               )
         ENDIF(CMAKE_CXX_COMPILER_ARG1)
     ELSE()
@@ -305,10 +305,13 @@ fi
 ${_command} '-D$<JOIN:$<TARGET_PROPERTY:${_targetName},COMPILE_DEFINITIONS>,' '-D>'
 ")
     GET_FILENAME_COMPONENT(_outdir ${_output} PATH)
+    if(NOT CMAKE_HOST_WIN32)  # chmod may be not available on Win32/MinGW (and it is not required)
+      set(_pch_prepare_command COMMAND chmod +x "${_pch_generate_file_cmd}")
+    endif()
     ADD_CUSTOM_COMMAND(
       OUTPUT "${_output}"
       COMMAND ${CMAKE_COMMAND} -E make_directory "${_outdir}"
-      COMMAND chmod +x "${_pch_generate_file_cmd}"
+      ${_pch_prepare_command}
       COMMAND "${_pch_generate_file_cmd}"
       DEPENDS "${_input}" "${_pch_generate_file_cmd}"
       DEPENDS "${CMAKE_CURRENT_BINARY_DIR}/${_name}"

cmake/OpenCVUtils.cmake

@@ -16,7 +16,8 @@ function(ocv_cmake_dump_vars)
     string(TOLOWER "${__variableName}" __variableName_lower)
     if((__variableName MATCHES "${regex}" OR __variableName_lower MATCHES "${regex_lower}")
         AND NOT __variableName_lower MATCHES "^__")
-      set(__VARS "${__VARS}${__variableName}=${${__variableName}}\n")
+      get_property(__value VARIABLE PROPERTY "${__variableName}")
+      set(__VARS "${__VARS}${__variableName}=${__value}\n")
     endif()
   endforeach()
   if(DUMP_TOFILE)
@@ -99,6 +100,30 @@ macro(ocv_update VAR)
   endif()
 endmacro()
 
+function(_ocv_access_removed_variable VAR ACCESS)
+  if(ACCESS STREQUAL "MODIFIED_ACCESS")
+    set(OPENCV_SUPPRESS_MESSAGE_REMOVED_VARIABLE_${VAR} 1 PARENT_SCOPE)
+    return()
+  endif()
+  if(ACCESS MATCHES "UNKNOWN_.*"
+      AND NOT OPENCV_SUPPRESS_MESSAGE_REMOVED_VARIABLE
+      AND NOT OPENCV_SUPPRESS_MESSAGE_REMOVED_VARIABLE_${VAR}
+  )
+    message(WARNING "OpenCV: Variable has been removed from CMake scripts: ${VAR}")
+    set(OPENCV_SUPPRESS_MESSAGE_REMOVED_VARIABLE_${VAR} 1 PARENT_SCOPE)  # suppress similar messages
+  endif()
+endfunction()
+macro(ocv_declare_removed_variable VAR)
+  if(NOT DEFINED ${VAR})  # don't hit external variables
+    variable_watch(${VAR} _ocv_access_removed_variable)
+  endif()
+endmacro()
+macro(ocv_declare_removed_variables)
+  foreach(_var ${ARGN})
+    ocv_declare_removed_variable(${_var})
+  endforeach()
+endmacro()
+
 # Search packages for the host system instead of packages for the target system
 # in case of cross compilation these macros should be defined by the toolchain file
 if(NOT COMMAND find_host_package)
@@ -287,9 +312,22 @@ function(ocv_append_target_property target prop)
   endif()
 endfunction()
 
+if(DEFINED OPENCV_DEPENDANT_TARGETS_LIST)
+  foreach(v ${OPENCV_DEPENDANT_TARGETS_LIST})
+    unset(${v} CACHE)
+  endforeach()
+  unset(OPENCV_DEPENDANT_TARGETS_LIST CACHE)
+endif()
+
 function(ocv_append_dependant_targets target)
   #ocv_debug_message("ocv_append_dependant_targets(${target} ${ARGN})")
   _ocv_fix_target(target)
+  list(FIND OPENCV_DEPENDANT_TARGETS_LIST "OPENCV_DEPENDANT_TARGETS_${target}" __id)
+  if(__id EQUAL -1)
+    list(APPEND OPENCV_DEPENDANT_TARGETS_LIST "OPENCV_DEPENDANT_TARGETS_${target}")
+    list(SORT OPENCV_DEPENDANT_TARGETS_LIST)
+    set(OPENCV_DEPENDANT_TARGETS_LIST "${OPENCV_DEPENDANT_TARGETS_LIST}" CACHE INTERNAL "")
+  endif()
   set(OPENCV_DEPENDANT_TARGETS_${target} "${OPENCV_DEPENDANT_TARGETS_${target}};${ARGN}" CACHE INTERNAL "" FORCE)
 endfunction()
 
@@ -363,6 +401,7 @@ macro(ocv_clear_vars)
 endmacro()
 
 set(OCV_COMPILER_FAIL_REGEX
+    "argument '.*' is not valid"                # GCC 9+
     "command line option .* is valid for .* but not for C\\+\\+" # GNU
     "command line option .* is valid for .* but not for C" # GNU
     "unrecognized .*option"                     # GNU
@@ -414,12 +453,34 @@ MACRO(ocv_check_compiler_flag LANG FLAG RESULT)
       else()
         set(__msg "")
       endif()
+      if(CMAKE_REQUIRED_LIBRARIES)
+        set(__link_libs LINK_LIBRARIES ${CMAKE_REQUIRED_LIBRARIES})
+      else()
+        set(__link_libs)
+      endif()
+      set(__cmake_flags "")
+      if(CMAKE_EXE_LINKER_FLAGS)  # CMP0056 do this on new CMake
+        list(APPEND __cmake_flags "-DCMAKE_EXE_LINKER_FLAGS=${CMAKE_EXE_LINKER_FLAGS}")
+      endif()
+
+      # CMP0067 do this on new CMake
+      if(DEFINED CMAKE_CXX_STANDARD)
+        list(APPEND __cmake_flags "-DCMAKE_CXX_STANDARD=${CMAKE_CXX_STANDARD}")
+      endif()
+      if(DEFINED CMAKE_CXX_STANDARD_REQUIRED)
+        list(APPEND __cmake_flags "-DCMAKE_CXX_STANDARD_REQUIRED=${CMAKE_CXX_STANDARD_REQUIRED}")
+      endif()
+      if(DEFINED CMAKE_CXX_EXTENSIONS)
+        list(APPEND __cmake_flags "-DCMAKE_CXX_EXTENSIONS=${CMAKE_CXX_EXTENSIONS}")
+      endif()
+
       MESSAGE(STATUS "Performing Test ${RESULT}${__msg}")
       TRY_COMPILE(${RESULT}
         "${CMAKE_BINARY_DIR}"
         "${_fname}"
-        CMAKE_FLAGS "-DCMAKE_EXE_LINKER_FLAGS=${CMAKE_EXE_LINKER_FLAGS}"   # CMP0056 do this on new CMake
+        CMAKE_FLAGS ${__cmake_flags}
         COMPILE_DEFINITIONS "${FLAG}"
+        ${__link_libs}
         OUTPUT_VARIABLE OUTPUT)
 
       if(${RESULT})
@@ -781,7 +842,7 @@ macro(ocv_check_modules define)
           if(pkgcfg_lib_${define}_${_lib})
             list(APPEND _libs "${pkgcfg_lib_${define}_${_lib}}")
           else()
-            message(WARNING "ocv_check_modules(${define}): can't find library '${_lib}'. Specify 'pkgcfg_lib_${define}_${_lib}' manualy")
+            message(WARNING "ocv_check_modules(${define}): can't find library '${_lib}'. Specify 'pkgcfg_lib_${define}_${_lib}' manually")
             list(APPEND _libs "${_lib}")
           endif()
         else()
@@ -1370,12 +1431,14 @@ endmacro()
 function(ocv_target_link_libraries target)
   set(LINK_DEPS ${ARGN})
   _ocv_fix_target(target)
-  set(LINK_MODE "LINK_PRIVATE")
+  set(LINK_MODE "PRIVATE")
   set(LINK_PENDING "")
   foreach(dep ${LINK_DEPS})
     if(" ${dep}" STREQUAL " ${target}")
       # prevent "link to itself" warning (world problem)
-    elseif(" ${dep}" STREQUAL " LINK_PRIVATE" OR " ${dep}" STREQUAL "LINK_PUBLIC")
+    elseif(" ${dep}" STREQUAL " LINK_PRIVATE" OR " ${dep}" STREQUAL " LINK_PUBLIC"  # deprecated
+        OR " ${dep}" STREQUAL " PRIVATE" OR " ${dep}" STREQUAL " PUBLIC" OR " ${dep}" STREQUAL " INTERFACE"
+    )
       if(NOT LINK_PENDING STREQUAL "")
         __ocv_push_target_link_libraries(${LINK_MODE} ${LINK_PENDING})
         set(LINK_PENDING "")
@@ -1514,7 +1577,10 @@ macro(ocv_get_all_libs _modules _extra _3rdparty)
         endif()
         if (TARGET ${dep})
           get_target_property(_type ${dep} TYPE)
-          if(_type STREQUAL "STATIC_LIBRARY" AND BUILD_SHARED_LIBS OR _type STREQUAL "INTERFACE_LIBRARY")
+          if((_type STREQUAL "STATIC_LIBRARY" AND BUILD_SHARED_LIBS)
+              OR _type STREQUAL "INTERFACE_LIBRARY"
+              OR DEFINED OPENCV_MODULE_${dep}_LOCATION  # OpenCV modules
+          )
             # nothing
           else()
             get_target_property(_output ${dep} IMPORTED_LOCATION)

cmake/android/android_ant_projects.cmake

@@ -49,7 +49,7 @@ macro(android_get_compatible_target VAR)
     list(GET ANDROID_SDK_TARGETS 0 __lvl)
     string(REGEX MATCH "[0-9]+$" __lvl "${__lvl}")
 
-    #find minimal level mathing to all provided levels
+    #find minimal level matching to all provided levels
     foreach(lvl ${ARGN})
       string(REGEX MATCH "[0-9]+$" __level "${lvl}")
       if(__level GREATER __lvl)
@@ -194,7 +194,7 @@ macro(add_android_project target path)
         add_library(${JNI_LIB_NAME} SHARED ${android_proj_jni_files})
         ocv_target_include_modules_recurse(${JNI_LIB_NAME} ${android_proj_NATIVE_DEPS})
         ocv_target_include_directories(${JNI_LIB_NAME} "${path}/jni")
-        ocv_target_link_libraries(${JNI_LIB_NAME} LINK_PRIVATE ${OPENCV_LINKER_LIBS} ${android_proj_NATIVE_DEPS})
+        ocv_target_link_libraries(${JNI_LIB_NAME} PRIVATE ${OPENCV_LINKER_LIBS} ${android_proj_NATIVE_DEPS})
 
         set_target_properties(${JNI_LIB_NAME} PROPERTIES
             OUTPUT_NAME "${JNI_LIB_NAME}"

cmake/checks/atomic_check.cpp

@@ -0,0 +1,12 @@
+#include <atomic>
+
+static int test()
+{
+    std::atomic<long long> x;
+    return x;
+}
+
+int main()
+{
+    return test();
+}

cmake/checks/cpu_avx2.cpp

@@ -6,5 +6,6 @@ void test()
 {
     int data[8] = {0,0,0,0, 0,0,0,0};
     __m256i a = _mm256_loadu_si256((const __m256i *)data);
+    __m256i b = _mm256_bslli_epi128(a, 1);  // available in GCC 4.9.3+
 }
 int main() { return 0; }

cmake/checks/cpu_avx512clx.cpp

@@ -3,9 +3,9 @@
 void test()
 {
     __m512i a, b, c;
-    a = _mm512_dpwssd_epi32(a, b, c);
+    a = _mm512_dpwssd_epi32(a, b, c);            // VNNI
 }
 #else
-#error "AVX512-CEL is not supported"
+#error "AVX512-CLX is not supported"
 #endif
-int main() { return 0; }
+int main() { return 0; }

cmake/checks/cpu_avx512icl.cpp

@@ -3,9 +3,10 @@
 void test()
 {
     __m512i a, b, c;
-    a = _mm512_popcnt_epi8(a);
-    a = _mm512_shrdv_epi64(a, b, c);
-    a = _mm512_popcnt_epi64(a);
+    a = _mm512_popcnt_epi8(a);                   // BITALG
+    a = _mm512_shrdv_epi64(a, b, c);             // VBMI2
+    a = _mm512_popcnt_epi64(a);                  // VPOPCNTDQ
+    a = _mm512_dpwssd_epi32(a, b, c);            // VNNI
 }
 #else
 #error "AVX512-ICL is not supported"

cmake/checks/cpu_msa.cpp

@@ -0,0 +1,23 @@
+#include <stdio.h>
+
+#if defined(__mips_msa)
+#  include <msa.h>
+#  define CV_MSA 1
+#endif
+
+#if defined CV_MSA
+int test()
+{
+    const float src[] = { 0.0f, 0.0f, 0.0f, 0.0f };
+    v4f32 val = (v4f32)__msa_ld_w((const float*)(src), 0);
+    return __msa_copy_s_w(__builtin_msa_ftint_s_w (val), 0);
+}
+#else
+#error "MSA is not supported"
+#endif
+
+int main()
+{
+  printf("%d\n", test());
+  return 0;
+}

cmake/checks/cpu_neon.cpp

@@ -1,6 +1,6 @@
 #include <stdio.h>
 
-#if defined _WIN32 && defined(_M_ARM)
+#if defined _WIN32 && (defined(_M_ARM) || defined(_M_ARM64))
 # include <Intrin.h>
 # include <arm_neon.h>
 # define CV_NEON 1
@@ -9,6 +9,10 @@
 #  define CV_NEON 1
 #endif
 
+// MSVC 2019 bug. Details: https://github.com/opencv/opencv/pull/16027
+void test_aliased_type(const uint8x16_t& a) { }
+void test_aliased_type(const int8x16_t& a) { }
+
 #if defined CV_NEON
 int test()
 {

cmake/platforms/OpenCV-Android.cmake

@@ -0,0 +1 @@
+# empty

cmake/platforms/OpenCV-Darwin.cmake

@@ -0,0 +1 @@
+# empty

cmake/platforms/OpenCV-Linux.cmake

@@ -0,0 +1 @@
+# empty

cmake/platforms/OpenCV-WinRT.cmake

@@ -0,0 +1,31 @@
+set(WINRT TRUE)
+
+add_definitions(-DWINRT -DNO_GETENV)
+
+# Making definitions available to other configurations and
+# to filter dependency restrictions at compile time.
+if(WINDOWS_PHONE)
+  set(WINRT_PHONE TRUE)
+  add_definitions(-DWINRT_PHONE)
+elseif(WINDOWS_STORE)
+  set(WINRT_STORE TRUE)
+  add_definitions(-DWINRT_STORE)
+endif()
+
+if(CMAKE_SYSTEM_VERSION MATCHES 10)
+  set(WINRT_10 TRUE)
+  add_definitions(-DWINRT_10)
+  add_definitions(/DWINVER=_WIN32_WINNT_WIN10 /DNTDDI_VERSION=NTDDI_WIN10 /D_WIN32_WINNT=_WIN32_WINNT_WIN10)
+elseif(CMAKE_SYSTEM_VERSION MATCHES 8.1)
+  set(WINRT_8_1 TRUE)
+  add_definitions(-DWINRT_8_1)
+  add_definitions(/DWINVER=_WIN32_WINNT_WINBLUE /DNTDDI_VERSION=NTDDI_WINBLUE /D_WIN32_WINNT=_WIN32_WINNT_WINBLUE)
+elseif(CMAKE_SYSTEM_VERSION MATCHES 8.0)
+  set(WINRT_8_0 TRUE)
+  add_definitions(-DWINRT_8_0)
+  add_definitions(/DWINVER=_WIN32_WINNT_WIN8 /DNTDDI_VERSION=NTDDI_WIN8 /D_WIN32_WINNT=_WIN32_WINNT_WIN8)
+else()
+  message(STATUS "Unsupported WINRT version (consider upgrading OpenCV): ${CMAKE_SYSTEM_VERSION}")
+endif()
+
+set(OPENCV_DEBUG_POSTFIX "")

cmake/platforms/OpenCV-Windows.cmake

@@ -0,0 +1 @@
+# empty

cmake/platforms/OpenCV-WindowsCE.cmake

@@ -0,0 +1 @@
+add_definitions(-DNO_GETENV)

cmake/platforms/OpenCV-WindowsPhone.cmake

@@ -0,0 +1,6 @@
+include("${CMAKE_CURRENT_LIST_DIR}/OpenCV-WinRT.cmake")
+
+# Adding additional using directory for WindowsPhone 8.0 to get Windows.winmd properly
+if(WINRT_8_0)
+  set(OPENCV_EXTRA_CXX_FLAGS "${OPENCV_EXTRA_CXX_FLAGS} /AI\$(WindowsSDK_MetadataPath)")
+endif()

cmake/platforms/OpenCV-WindowsStore.cmake

@@ -0,0 +1 @@
+include("${CMAKE_CURRENT_LIST_DIR}/OpenCV-WinRT.cmake")

cmake/platforms/OpenCV-iOS.cmake

@@ -0,0 +1 @@
+# empty

cmake/templates/OpenCVConfig.root-WIN32.cmake.in

@@ -84,17 +84,31 @@ endfunction()
 
 get_filename_component(OpenCV_CONFIG_PATH "${CMAKE_CURRENT_LIST_FILE}" DIRECTORY)
 
+if((NOT DEFINED CMAKE_SYSTEM_PROCESSOR OR CMAKE_SYSTEM_PROCESSOR STREQUAL "")
+    AND NOT OPENCV_SUPPRESS_MESSAGE_MISSING_CMAKE_SYSTEM_PROCESSOR)
+  message(WARNING "OpenCV: CMAKE_SYSTEM_PROCESSOR is not defined. Perhaps CMake toolchain is broken")
+endif()
+if(NOT DEFINED CMAKE_SIZEOF_VOID_P
+    AND NOT OPENCV_SUPPRESS_MESSAGE_MISSING_CMAKE_SIZEOF_VOID_P)
+  message(WARNING "OpenCV: CMAKE_SIZEOF_VOID_P is not defined. Perhaps CMake toolchain is broken")
+endif()
+
 if(DEFINED OpenCV_ARCH AND DEFINED OpenCV_RUNTIME)
   # custom overridden values
 elseif(MSVC)
-  if(CMAKE_CL_64)
-    set(OpenCV_ARCH x64)
-  elseif((CMAKE_GENERATOR MATCHES "ARM") OR ("${arch_hint}" STREQUAL "ARM") OR (CMAKE_VS_EFFECTIVE_PLATFORMS MATCHES "ARM|arm"))
-    # see Modules/CmakeGenericSystem.cmake
-    set(OpenCV_ARCH ARM)
+  # see Modules/CMakeGenericSystem.cmake
+  if("${CMAKE_GENERATOR}" MATCHES "(Win64|IA64)")
+    set(OpenCV_ARCH "x64")
+  elseif("${CMAKE_GENERATOR_PLATFORM}" MATCHES "ARM64")
+    set(OpenCV_ARCH "ARM64")
+  elseif("${CMAKE_GENERATOR}" MATCHES "ARM")
+    set(OpenCV_ARCH "ARM")
+  elseif("${CMAKE_SIZEOF_VOID_P}" STREQUAL "8")
+    set(OpenCV_ARCH "x64")
   else()
     set(OpenCV_ARCH x86)
   endif()
+
   if(MSVC_VERSION EQUAL 1400)
     set(OpenCV_RUNTIME vc8)
   elseif(MSVC_VERSION EQUAL 1500)
@@ -127,11 +141,7 @@ elseif(MSVC)
 elseif(MINGW)
   set(OpenCV_RUNTIME mingw)
 
-  execute_process(COMMAND ${CMAKE_CXX_COMPILER} -dumpmachine
-                  OUTPUT_VARIABLE OPENCV_GCC_TARGET_MACHINE
-                  OUTPUT_STRIP_TRAILING_WHITESPACE)
-  if(OPENCV_GCC_TARGET_MACHINE MATCHES "amd64|x86_64|AMD64")
-    set(MINGW64 1)
+  if(CMAKE_SYSTEM_PROCESSOR MATCHES "amd64.*|x86_64.*|AMD64.*")
     set(OpenCV_ARCH x64)
   else()
     set(OpenCV_ARCH x86)

cmake/templates/dllmain.cpp.in

@@ -18,11 +18,17 @@ namespace cv {
 extern __declspec(dllimport) bool __termination;  // Details: #12750
 }
 
-extern "C"
-BOOL WINAPI DllMain(HINSTANCE, DWORD fdwReason, LPVOID lpReserved);
+#ifdef _WIN32_WCE
+#define DLL_MAIN_ARG0 HANDLE
+#else
+#define DLL_MAIN_ARG0 HINSTANCE
+#endif
 
 extern "C"
-BOOL WINAPI DllMain(HINSTANCE, DWORD fdwReason, LPVOID lpReserved)
+BOOL WINAPI DllMain(DLL_MAIN_ARG0, DWORD fdwReason, LPVOID lpReserved);
+
+extern "C"
+BOOL WINAPI DllMain(DLL_MAIN_ARG0, DWORD fdwReason, LPVOID lpReserved)
 {
     if (fdwReason == DLL_THREAD_DETACH || fdwReason == DLL_PROCESS_DETACH)
     {

cmake/templates/opencv_abi.xml.in

@@ -22,6 +22,8 @@
 
 <skip_headers>
     opencv2/core/hal/intrin*
+    opencv2/core/hal/*macros.*
+    opencv2/core/hal/*.impl.*
     opencv2/core/cuda*
     opencv2/core/opencl*
     opencv2/core/private*

doc/CMakeLists.txt

@@ -147,7 +147,7 @@ if(DOXYGEN_FOUND)
 
   # set export variables
   string(REPLACE ";" " \\\n" CMAKE_DOXYGEN_INPUT_LIST "${rootfile} ; ${faqfile} ; ${paths_include} ; ${paths_hal_interface} ; ${paths_doc} ; ${tutorial_path} ; ${tutorial_py_path} ; ${tutorial_js_path} ; ${paths_tutorial} ; ${tutorial_contrib_root}")
-  string(REPLACE ";" " \\\n" CMAKE_DOXYGEN_IMAGE_PATH "${paths_doc} ; ${tutorial_path} ; ${tutorial_py_path} ; ${tutorial_js_path} ; ${paths_tutorial}")
+  string(REPLACE ";" " \\\n" CMAKE_DOXYGEN_IMAGE_PATH "${CMAKE_CURRENT_SOURCE_DIR}/images ; ${paths_doc} ; ${tutorial_path} ; ${tutorial_py_path} ; ${tutorial_js_path} ; ${paths_tutorial}")
   string(REPLACE ";" " \\\n" CMAKE_DOXYGEN_EXCLUDE_LIST "${CMAKE_DOXYGEN_EXCLUDE_LIST}")
   string(REPLACE ";" " " CMAKE_DOXYGEN_ENABLED_SECTIONS "${CMAKE_DOXYGEN_ENABLED_SECTIONS}")
   # TODO: remove paths_doc from EXAMPLE_PATH after face module tutorials/samples moved to separate folders
@@ -172,9 +172,27 @@ if(DOXYGEN_FOUND)
   list(APPEND CMAKE_DOXYGEN_HTML_FILES "${CMAKE_CURRENT_SOURCE_DIR}/tutorial-utils.js")
   string(REPLACE ";" " \\\n" CMAKE_DOXYGEN_HTML_FILES "${CMAKE_DOXYGEN_HTML_FILES}")
 
+  set(OPENCV_DOCS_DOT_PATH "" CACHE PATH "Doxygen/DOT_PATH value")
+  set(CMAKECONFIG_DOT_PATH "${OPENCV_DOCS_DOT_PATH}")
+
+  set(OPENCV_DOCS_HAVE_DOT "NO" CACHE BOOL "Doxygen: build extra diagrams")
+  set(CMAKECONFIG_HAVE_DOT "${OPENCV_DOCS_HAVE_DOT}")
+
+  # 'png' is good enough for compatibility (but requires +50% storage space)
+  set(OPENCV_DOCS_DOT_IMAGE_FORMAT "svg" CACHE STRING "Doxygen/DOT_IMAGE_FORMAT value")
+  set(CMAKECONFIG_DOT_IMAGE_FORMAT "${OPENCV_DOCS_DOT_IMAGE_FORMAT}")
+
+  # Doxygen 1.8.16 fix: https://github.com/doxygen/doxygen/pull/6870
+  # NO is needed here: https://github.com/opencv/opencv/pull/16039
+  set(OPENCV_DOCS_INTERACTIVE_SVG "NO" CACHE BOOL "Doxygen/INTERACTIVE_SVG value")
+  set(CMAKECONFIG_INTERACTIVE_SVG "${OPENCV_DOCS_INTERACTIVE_SVG}")
+
+  set(OPENCV_DOCS_DOXYFILE_IN "Doxyfile.in" CACHE PATH "Doxygen configuration template file (Doxyfile.in)")
+  set(OPENCV_DOCS_DOXYGEN_LAYOUT "DoxygenLayout.xml" CACHE PATH "Doxygen layout file (.xml)")
+
   # writing file
-  configure_file(DoxygenLayout.xml DoxygenLayout.xml @ONLY)
-  configure_file(Doxyfile.in ${doxyfile} @ONLY)
+  configure_file("${OPENCV_DOCS_DOXYGEN_LAYOUT}" DoxygenLayout.xml @ONLY)
+  configure_file("${OPENCV_DOCS_DOXYFILE_IN}" ${doxyfile} @ONLY)
   configure_file(root.markdown.in ${rootfile} @ONLY)
 
   # js tutorial assets

doc/Doxyfile.in

@@ -266,12 +266,10 @@ GENERATE_TAGFILE       = @CMAKE_DOXYGEN_OUTPUT_PATH@/html/opencv.tag
 ALLEXTERNALS           = NO
 EXTERNAL_GROUPS        = YES
 EXTERNAL_PAGES         = YES
-PERL_PATH              = /usr/bin/perl
 CLASS_DIAGRAMS         = YES
-MSCGEN_PATH            =
 DIA_PATH               =
 HIDE_UNDOC_RELATIONS   = NO
-HAVE_DOT               = NO
+HAVE_DOT               = @CMAKECONFIG_HAVE_DOT@
 DOT_NUM_THREADS        = 0
 DOT_FONTNAME           = Helvetica
 DOT_FONTSIZE           = 10
@@ -288,9 +286,9 @@ CALL_GRAPH             = YES
 CALLER_GRAPH           = NO
 GRAPHICAL_HIERARCHY    = YES
 DIRECTORY_GRAPH        = YES
-DOT_IMAGE_FORMAT       = svg
-INTERACTIVE_SVG        = YES
-DOT_PATH               =
+DOT_IMAGE_FORMAT       = @CMAKECONFIG_DOT_IMAGE_FORMAT@
+INTERACTIVE_SVG        = @CMAKECONFIG_INTERACTIVE_SVG@
+DOT_PATH               = @CMAKECONFIG_DOT_PATH@
 DOTFILE_DIRS           =
 MSCFILE_DIRS           =
 DIAFILE_DIRS           =