Control: reassign -1 gcc-6 6-20160117-1 Control: affects -1 ffmpeg Hi Martin,
On 20.01.2016 04:44, Martin Michlmayr wrote: > This package fails to build with GCC 6. GCC 6 has not been released > yet, but it's expected that GCC 6 will become the default compiler for > stretch. > > Note that only the first error is reported; there might be more. You > can find a snapshot of GCC 6 in experimental. To build with GCC 6, > you can set CC=gcc-6 CXX=g++-6 explicitly. > > You may be able to find out more about this issue at > https://gcc.gnu.org/gcc-6/changes.html > >> sbuild (Debian sbuild) 0.67.0 (26 Dec 2015) on dl580gen9-02.hlinux > ... >> cc -I. -I/<<PKGBUILDDIR>>/ -Wdate-time -D_FORTIFY_SOURCE=2 -U_FORTIFY_SOURCE >> -D_FORTIFY_SOURCE=2 -D_ISOC99_SOURCE -D_FILE_OFFSET_BITS=64 >> -D_LARGEFILE_SOURCE -D_POSIX_C_SOURCE=200112 -D_XOPEN_SOURCE=600 -DPIC >> -DZLIB_CONST -g -O2 -fstack-protector-strong -Wformat >> -Werror=format-security -fno-strict-overflow -fstack-protector-all >> -std=c99 -fomit-frame-pointer -fPIC -pthread -I/usr/include/p11-kit-1 >> -I/usr/include/harfbuzz -I/usr/include/freetype2 -I/usr/include/fribidi >> -I/usr/include/freetype2 -I/usr/include/bs2b -I/usr/include/freetype2 >> -I/usr/include/freetype2 -I/usr/include/fribidi -I/usr/include/opencv >> -I/usr/include/opus -D_REENTRANT -I/usr/include/p11-kit-1 >> -I/usr/include/schroedinger-1.0 -I/usr/include/orc-0.4 >> -D_GNU_SOURCE=1 -D_REENTRANT -I/usr/include/SDL -g >> -Wdeclaration-after-statement -Wall -Wdisabled-optimization -Wpointer-arith >> -Wredundant-decls -Wwrite-strings -Wtype-limits -Wundef -Wmissing-prototypes >> -Wno-pointer-to-int-cast -Wstrict-prototypes >> -Wempty-body -Wno-parentheses -Wno-switch -Wno-format-zero-length >> -Wno-pointer-sign -O3 -fno-math-errno -fno-signed-zeros -fno-tree-vectorize >> -Werror=format-security -Werror=implicit-function-declaration >> -Werror=missing-prototypes -Werror=return-type -Werror=vla -Wformat >> -fdiagnostics-color=auto -Wno-maybe-uninitialized -c -o ffmpeg_opt.o >> /<<PKGBUILDDIR>>/ffmpeg_opt.c >> /<<PKGBUILDDIR>>/ffplay.c:3217:46: error: missing binary operator before >> token "(" >> #if defined(__APPLE__) && SDL_VERSION_ATLEAST(1, 2, 14) >> ^ >> cc -I. -I/<<PKGBUILDDIR>>/ -Wdate-time -D_FORTIFY_SOURCE=2 -U_FORTIFY_SOURCE >> -D_FORTIFY_SOURCE=2 -D_ISOC99_SOURCE -D_FILE_OFFSET_BITS=64 >> -D_LARGEFILE_SOURCE -D_POSIX_C_SOURCE=200112 -D_XOPEN_SOURCE=600 -DPIC >> -DZLIB_CONST -DHAVE_AV_CONFIG_H -MM >> /<<PKGBUILDDIR>>/libavfilter/avf_showvolume.c | sed -e "/^#.*/d" -e >> "s,^[[:space:]]*avf_showvolume\\.o,libavfilter/avf_showvolume.o," > >> libavfilter/avf_showvolume.d >> /<<PKGBUILDDIR>>/libavfilter/avf_showcqt.c:38:10: error: #include expects >> "FILENAME" or <FILENAME> >> #include FT_FREETYPE_H >> ^~~~~~~~~~~~~ > ... >> -Wstrict-prototypes -Wempty-body -Wno-parentheses -Wno-switch >> -Wno-format-zero-length -Wno-pointer-sign -O3 -fno-math-errno >> -fno-signed-zeros -fno-tree-vectorize -Werror=format-security >> -Werror=implicit-function-declaration -Werror=missing-prototypes >> -Werror=return-type -Werror=vla -Wformat -fdiagnostics-color=auto >> -Wno-maybe-uninitialized -c -o libavfilter/vf_drawbox.o >> /<<PKGBUILDDIR>>/libavfilter/vf_drawbox.c >> /<<PKGBUILDDIR>>/libavfilter/vf_drawtext.c:69:10: error: #include expects >> "FILENAME" or <FILENAME> >> #include FT_FREETYPE_H >> ^~~~~~~~~~~~~ >> >> /<<PKGBUILDDIR>>/libavfilter/vf_drawtext.c:70:10: error: #include expects >> "FILENAME" or <FILENAME> >> #include FT_GLYPH_H >> ^~~~~~~~~~ >> >> /<<PKGBUILDDIR>>/libavfilter/vf_drawtext.c:71:10: error: #include expects >> "FILENAME" or <FILENAME> >> #include FT_STROKER_H >> ^~~~~~~~~~~~ >> >> /<<PKGBUILDDIR>>/libavfilter/vf_drawtext.c:275:10: error: #include expects >> "FILENAME" or <FILENAME> >> #include FT_ERRORS_H >> ^~~~~~~~~~~ These "errors" are a red herring: they happen during dependency generation and are thus harmless. They can also be seen when building in unstable. (They now get suppressed upstream.) The real problems are test failures: [...] Test vsynth1-svq1 failed. Look at tests/data/fate/vsynth1-svq1.err for details. [...] Test vsynth3-svq1 failed. Look at tests/data/fate/vsynth3-svq1.err for details. [...] Test vsynth2-svq1 failed. Look at tests/data/fate/vsynth2-svq1.err for details. These worked fine when I tested building with gcc-6 6-20151211-1. So this seems to be a regression introduced during the last month. Interestingly these tests pass upstream [1] when building with "gcc 6.0.0 (GCC) 20160119 (experimental)". So this either got fixed between 17. and 19. January, or it is a Debian specific problem. The problem is in svq1_encode_plane in libavcodec/svq1enc.c. For P-frames, the decoded_plane buffer gets overwritten with wrong values: $ cd debian/standard $ export LD_LIBRARY_PATH="libavcodec:libavdevice:libavfilter:libavformat:libavresample:libavutil:libpostproc:libswresample:libswscale" $ gdb --args ./ffmpeg -nostdin -nostats -cpuflags all -f rawvideo -s 352x288 -pix_fmt yuv420p -threads 1 -idct simple -flags +bitexact -sws_flags +accurate_rnd+bitexact -fflags +bitexact -hwaccel none -threads 1 -thread_type frame+slice -i ./tests/data/vsynth1.yuv -threads 1 -idct simple -dct fastint -c svq1 -qscale 3 -pix_fmt yuv410p -flags +bitexact -sws_flags +accurate_rnd+bitexact -fflags +bitexact -f framecrc - GNU gdb (Debian 7.10-1+b1) 7.10 [...] (gdb) br svq1_encode_plane [...] (gdb) r [...] ffmpeg version 2.8.5-1 Copyright (c) 2000-2016 the FFmpeg developers built with gcc 6.0.0 (Debian 6-20160117-1) 20160117 (experimental) [trunk revision 232481] [...] Breakpoint 1, svq1_encode_plane (s=s@entry=0x697620, plane=plane@entry=0, src_plane=0x6deb00 "[...]"..., ref_plane=0x6fae80 "", decoded_plane=0x71cb60 "", width=352, height=288, src_stride=352, stride=384) at libavcodec/svq1enc.c:248 248 { (gdb) c Continuing. Breakpoint 1, svq1_encode_plane (s=s@entry=0x697620, plane=plane@entry=1, src_plane=0x6f7780 "[...]", ref_plane=0x718ee0 "", decoded_plane=0x73aba0 "", width=88, height=72, src_stride=96, stride=96) at libavcodec/svq1enc.c:248 248 { (gdb) Continuing. Breakpoint 1, svq1_encode_plane (s=s@entry=0x697620, plane=plane@entry=2, src_plane=0x6f9300 "[...]", ref_plane=0x71ad20 "", decoded_plane=0x73c9e0 "", width=88, height=72, src_stride=96, stride=96) at libavcodec/svq1enc.c:248 248 { (gdb) Continuing. 0, 0, 0, 1, 36508, 0xb057df97, S=1, 8, 0x031b0064 Breakpoint 1, svq1_encode_plane (s=s@entry=0x697620, plane=plane@entry=0, src_plane=0x768dc0 "[...]"..., decoded_plane=0x6fae80 "", width=352, height=288, src_stride=352, stride=384) at libavcodec/svq1enc.c:248 248 { (gdb) advance 456 svq1_encode_plane (s=s@entry=0x697620, plane=plane@entry=0, src_plane=0x768dc0 "[...]"..., ref_plane=0x71cb60 "[...]"..., decoded_plane=0x6fae80 "[...]", width=352, height=288, src_stride=352, stride=384) at libavcodec/svq1enc.c:456 456 if (score[2] < score[best] && mx == 0 && my == 0) { (gdb) p score $1 = {2147483647, 26545, 0, 0} (gdb) # best = score[1] <= score[0]; (gdb) p best $2 = 1 (gdb) n 34 480 s->rd_total += score[best]; (gdb) p s->rd_total $3 = 7095940 (gdb) p best $4 = <optimized out> (gdb) n 463 if (best == 1) { (gdb) p s->rd_total - 7095940 $5 = 26545 (gdb) p score $6 = {2147483647, 26545, 624803, 0} (gdb) # So best was still 1 on line 480, as it should be. (gdb) p best $7 = 0 (gdb) # This is wrong, nothing should have changed best. (gdb) n 484 avpriv_copy_bits(&s->pb, reorder_buffer[best][i], (gdb) 483 for (i = 5; i >= 0; i--) (gdb) 483 for (i = 5; i >= 0; i--) (gdb) 483 for (i = 5; i >= 0; i--) (gdb) 483 for (i = 5; i >= 0; i--) (gdb) 483 for (i = 5; i >= 0; i--) (gdb) 483 for (i = 5; i >= 0; i--) (gdb) 487 s->hdsp.put_pixels_tab[0][0](decoded, temp, stride, 16); (gdb) # This only happens if (best == 0), which should not be the case here, but: (gdb) p best $8 = 0 (gdb) # Thus wrong values get written to decoded_plane: (gdb) p/x *(uint64_t*)decoded_plane@2 $9 = {0x756f68615c55524d, 0xa09d9a9691878079} (gdb) n 375 for (x = 0; x < block_width; x++) { (gdb) p/x *(uint64_t*)decoded_plane@2 $10 = {0x8ac9d9d8d6d6d6d6, 0x8b8b8a898a8b8a8a} As this is a compiler bug in gcc-6, I'm reassigning accordingly. Best regards, Andreas 1: http://fate.ffmpeg.org/report.cgi?time=20160119133541&slot=x86_64-archlinux-gcc-experimental
