Created attachment 791123 [details] The complete build log mkdir -p `dirname 64/gimp-texture.png` && GEGL_USE_OPENCL=no GEGL_SWAP=ram /usr/bin/gegl-0.4 ../../icons/Symbolic/64/gimp-texture.png -o 64/gimp-texture.png -- cast-format input-format="R'G'B'A float" output-format="RGBA float" brightness-contrast contrast=1.5 cast-format input-format="RGBA float" output-format="R'G'B'A float" make[3]: *** [Makefile:2446: 64/gimp-texture.png] Segmentation fault make[3]: Leaving directory '/tmp/portage/media-gfx/gimp-2.10.32/work/gimp-2.10.32/icons/Symbolic-High-Contrast' make[2]: *** [Makefile:671: all-recursive] Error 1 make[2]: Leaving directory '/tmp/portage/media-gfx/gimp-2.10.32/work/gimp-2.10.32/icons' make[1]: *** [Makefile:862: all-recursive] Error 1 make[1]: Leaving directory '/tmp/portage/media-gfx/gimp-2.10.32/work/gimp-2.10.32' make: *** [Makefile:763: all] Error 2 > emerge --info '=media-gfx/gimp-2.10.32::gentoo' Portage 3.0.30 (python 3.10.5-final-0, default/linux/amd64/17.1/desktop, gcc-11.3.0, glibc-2.34-r13, 5.18.10-rc1-alb x86_64) ================================================================= System Settings ================================================================= System uname: Linux-5.18.10-rc1-alb-x86_64-Dual_Core_AMD_Opteron-tm-_Processor_290-with-glibc2.34 KiB Mem: 16489868 total, 1144344 free KiB Swap: 67108852 total, 61889232 free Timestamp of repository gentoo: Mon, 11 Jul 2022 18:15:01 +0000 Head commit of repository gentoo: cccacbae494c2783b04aefa08d0c55295343c8f5 sh bash 5.1_p16 ld GNU ld (Gentoo 2.37_p1 p2) 2.37 app-misc/pax-utils: 1.3.4::gentoo app-shells/bash: 5.1_p16::gentoo dev-java/java-config: 2.3.1::gentoo dev-lang/perl: 5.34.1-r3::gentoo dev-lang/python: 3.9.13::gentoo, 3.10.5::gentoo dev-lang/rust: 1.60.0::gentoo dev-util/cmake: 3.22.4::gentoo dev-util/meson: 0.62.2::gentoo sys-apps/baselayout: 2.8::gentoo sys-apps/openrc: 0.44.10::gentoo sys-apps/sandbox: 2.29::gentoo sys-devel/autoconf: 2.13-r2::gentoo, 2.71-r1::gentoo sys-devel/automake: 1.16.5::gentoo sys-devel/binutils: 2.37_p1-r2::gentoo sys-devel/binutils-config: 5.4.1::gentoo sys-devel/clang: 14.0.4::gentoo sys-devel/gcc: 11.3.0::gentoo sys-devel/gcc-config: 2.5-r1::gentoo sys-devel/libtool: 2.4.7::gentoo sys-devel/llvm: 14.0.4::gentoo sys-devel/make: 4.3::x-portage sys-kernel/linux-headers: 5.15-r3::gentoo (virtual/os-headers) sys-libs/glibc: 2.34-r13::gentoo Repositories: gentoo location: /usr/portage sync-type: rsync sync-uri: rsync://rsync.europe.gentoo.org/gentoo-portage priority: -1000 sync-rsync-verify-max-age: 24 sync-rsync-verify-jobs: 1 sync-rsync-verify-metamanifest: yes sync-rsync-extra-opts: x-portage location: /usr/local/portage masters: gentoo priority: 0 guru location: /var/lib/layman/guru masters: gentoo priority: 50 ACCEPT_KEYWORDS="amd64" ACCEPT_LICENSE="@FREE" CBUILD="x86_64-pc-linux-gnu" CFLAGS="-march=native -mtune=native -O2 -pipe -fomit-frame-pointer -finline-functions-called-once -ftree-vectorize" CHOST="x86_64-pc-linux-gnu" CONFIG_PROTECT="/etc /etc/stunnel/stunnel.conf /usr/lib64/libreoffice/program/sofficerc /usr/share/config /usr/share/gnupg/qualified.txt /usr/share/themes/oxygen-gtk/gtk-2.0 /var/lib/i2pd/certificates" CONFIG_PROTECT_MASK="/etc/ca-certificates.conf /etc/dconf /etc/env.d /etc/fonts/fonts.conf /etc/gconf /etc/gentoo-release /etc/php/apache2-php7.4/ext-active/ /etc/php/apache2-php8.1/ext-active/ /etc/php/cgi-php7.4/ext-active/ /etc/php/cgi-php8.1/ext-active/ /etc/php/cli-php7.4/ext-active/ /etc/php/cli-php8.1/ext-active/ /etc/php/fpm-php7.4/ext-active/ /etc/php/fpm-php8.1/ext-active/ /etc/php/phpdbg-php7.4/ext-active/ /etc/php/phpdbg-php8.1/ext-active/ /etc/revdep-rebuild /etc/sandbox.d /etc/terminfo /etc/texmf/language.dat.d /etc/texmf/language.def.d /etc/texmf/updmap.d /etc/texmf/web2c" CXXFLAGS="-march=native -mtune=native -O2 -pipe -fomit-frame-pointer -finline-functions-called-once -ftree-vectorize" DISTDIR="/scratch/portage/distfiles" ENV_UNSET="CARGO_HOME DBUS_SESSION_BUS_ADDRESS DISPLAY GOBIN GOPATH PERL5LIB PERL5OPT PERLPREFIX PERL_CORE PERL_MB_OPT PERL_MM_OPT XAUTHORITY XDG_CACHE_HOME XDG_CONFIG_HOME XDG_DATA_HOME XDG_RUNTIME_DIR" FCFLAGS="-O2 -pipe" FEATURES="assume-digests binpkg-docompress binpkg-dostrip binpkg-logs buildpkg-live config-protect-if-modified distlocks ebuild-locks fixlafiles ipc-sandbox merge-sync multilib-strict network-sandbox news parallel-fetch pid-sandbox preserve-libs protect-owned qa-unresolved-soname-deps sandbox sfperms strict unknown-features-warn unmerge-logs unmerge-orphans userfetch userpriv usersandbox usersync xattr" FFLAGS="-march=native -mtune=native -O2 -pipe -fomit-frame-pointer -finline-functions-called-once -ftree-vectorize -fprefetch-loop-arrays -funroll-loops -fno-stack-protector" GENTOO_MIRRORS="https://mirror.netcologne.de/gentoo/" LANG="en_US.UTF-8" LC_ALL="" LDFLAGS="-Wl,-O1 -Wl,--as-needed" LINGUAS="en fa ru" MAKEOPTS="-j3" PKGDIR="/var/cache/binpkgs" PORTAGE_CONFIGROOT="/" PORTAGE_RSYNC_OPTS="--recursive --links --safe-links --perms --times --omit-dir-times --compress --force --whole-file --delete --stats --human-readable --timeout=180 --exclude=/distfiles --exclude=/local --exclude=/packages --exclude=/.git" PORTAGE_TMPDIR="/tmp" SHELL="/bin/bash" USE="X a52 aac aacs acl acpi alsa amd64 amr apng appindicator argyllcms asm bash-completion bidi bluetooth bluray branding bundled-libs bzip2 cairo caps cdda cdr celt cjk cli colord crypt cryptsetup cscope cups dbus dirac djvu dri dts dv dvb dvd dvdr dvdread elogind encode eselect exif faac faad ffmpeg flac fontconfig fortran g726 g729 gdbm gif gimp gmp gpm gsm gsm-nonstandard gtk gtk3 gui http iconv icu idn ieee1394 ilbc jpeg jpeg2k jumbo-build kerberos lame lcms ldap ldapsam libglvnd libnotify libtirpc lm_sensors lock logrotate mad matroska mmap mms mng mp3 mp4 mpeg multilib musepack ncurses nls nodrm nptl nsplugin numa ogg opencl opengl openmp opus pam pango pcre pdf pkcs11 png policykit ppds qt5 raw readline samba scanner sdl seccomp session silk smartcard spell split-usr srtp ssl startup-notification svg taglib tcpd theora threads tiff timidity truetype udev udisks unicode upower usb vcd vdpau vhosts vim-syntax visio vorbis vpx wavpack webp winbind wireguard wmf wpg wxwidgets x264 xattr xcb xcomposite xinerama xml xmp xv xvid xvmc zlib" ABI_X86="32 64" ADA_TARGET="gnat_2020" APACHE2_MODULES="alias auth_basic auth_digest authn_alias authn_dbm authn_default authn_file authz_dbm authz_default authz_groupfile authz_host authz_user autoindex dir env expires filter headers deflate info log_config logio mime mime_magic negotiation status unique_id userdir rewrite reqtimeout proxy proxy_connect proxy_http authn_core authz_core unixd socache_shmcb" CALLIGRA_FEATURES="karbon sheets words" COLLECTD_PLUGINS="df interface irq load memory rrdtool swap syslog" CPU_FLAGS_X86="3dnow 3dnowext mmx mmxext sse sse2 sse3" ELIBC="glibc" GPSD_PROTOCOLS="ashtech aivdm earthmate evermore fv18 garmin garmintxt gpsclock greis isync itrax mtk3301 nmea ntrip navcom oceanserver oldstyle oncore rtcm104v2 rtcm104v3 sirf skytraq superstar2 timing tsip tripmate tnt ublox ubx" GRUB_PLATFORMS="pc" INPUT_DEVICES="evdev wacom" KERNEL="linux" L10N="en fa ru" LCD_DEVICES="bayrad cfontz cfontz633 glk hd44780 lb216 lcdm001 mtxorb ncurses text" LIBREOFFICE_EXTENSIONS="presenter-console presenter-minimizer" LUA_SINGLE_TARGET="lua5-1" LUA_TARGETS="lua5-1" NGINX_MODULES_HTTP="auth_pam access auth_basic autoindex browser charset fastcgi addition fancyindex geoip gzip headers_more limit_conn limit_req proxy referer rewrite scgi stub_status vhost_traffic_status sub" OFFICE_IMPLEMENTATION="libreoffice" PHP_TARGETS="php7-4 php8-0" POSTGRES_TARGETS="postgres12 postgres13" PYTHON_SINGLE_TARGET="python3_10" PYTHON_TARGETS="python3_10" RUBY_TARGETS="ruby27" SANE_BACKENDS="mustek net" USERLAND="GNU" VIDEO_CARDS="nvidia" XTABLES_ADDONS="quota2 psd pknock lscan length2 ipv4options ipset ipp2p iface geoip fuzzy condition tee tarpit sysrq proto steal rawnat logmark ipmark dhcpmac delude chaos account" Unset: ADDR2LINE, AR, ARFLAGS, AS, ASFLAGS, CC, CCLD, CONFIG_SHELL, CPP, CPPFLAGS, CTARGET, CXX, CXXFILT, ELFEDIT, EMERGE_DEFAULT_OPTS, EXTRA_ECONF, F77FLAGS, FC, GCOV, GPROF, INSTALL_MASK, LD, LEX, LFLAGS, LIBTOOL, MAKE, MAKEFLAGS, NM, OBJCOPY, OBJDUMP, PORTAGE_BINHOST, PORTAGE_BUNZIP2_COMMAND, PORTAGE_COMPRESS, PORTAGE_COMPRESS_FLAGS, PORTAGE_RSYNC_EXTRA_OPTS, RANLIB, READELF, RUSTFLAGS, SIZE, STRINGS, STRIP, YACC, YFLAGS
Could you, please, inform is this issue only for gimp-2.10.32 or was observed early on you system (e.g. gimp-2.10.28 or 2.10.30)?
Here is the backtrace: Thread 1 "gegl-0.4" received signal SIGSEGV, Segmentation fault. universal_nonlinear_rgb_u8_converter_sse2 (conversion=<optimized out>, src_char=0x5555559479a0 "\343\340\366[\252\326\372\\\240", dst_char=0x555555947900 "", samples=32, data=0x5555558df040) at ../babl-0.1.92/babl/base/babl-rgb-converter.c:318 318 ../babl-0.1.92/babl/base/babl-rgb-converter.c: No such file or directory. (gdb) bt #0 universal_nonlinear_rgb_u8_converter_sse2 (conversion=<optimized out>, src_char=0x5555559479a0 "\343\340\366[\252\326\372\\\240", dst_char=0x555555947900 "", samples=32, data=0x5555558df040) at ../babl-0.1.92/babl/base/babl-rgb-converter.c:318 #1 0x00007ffff7d1373f in _babl_process (n=32, destination=destination@entry=0x3e8, source=source@entry=0x3e8, cbabl=<optimized out>) at ../babl-0.1.92/babl/babl-fish-path.c:1323 #2 babl_process (babl=<optimized out>, source=source@entry=0x5555559479a0, destination=destination@entry=0x555555947900, n=n@entry=32) at ../babl-0.1.92/babl/babl-fish-path.c:1333 #3 0x00007ffff7d0f217 in babl_conversion_error (conversion=0x5555555f0510) at ../babl-0.1.92/babl/babl-conversion.c:504 #4 0x00007ffff7d122a8 in get_conversion_path (pc=pc@entry=0x7fffffffc870, current_format=current_format@entry=0x555555598e40, current_length=current_length@entry=3, max_length=max_length@entry=3, legal_error=legal_error@entry=4.6999999999999999e-06) at ../babl-0.1.92/babl/babl-fish-path.c:690 #5 0x00007ffff7d12268 in get_conversion_path (pc=pc@entry=0x7fffffffc870, current_format=current_format@entry=0x5555555964f0, current_length=current_length@entry=2, max_length=max_length@entry=3, legal_error=legal_error@entry=4.6999999999999999e-06) at ../babl-0.1.92/babl/babl-fish-path.c:749 #6 0x00007ffff7d12268 in get_conversion_path (pc=pc@entry=0x7fffffffc870, current_format=current_format@entry=0x555555613ea0, current_length=current_length@entry=1, max_length=max_length@entry=3, legal_error=legal_error@entry=4.6999999999999999e-06) at ../babl-0.1.92/babl/babl-fish-path.c:749 #7 0x00007ffff7d12268 in get_conversion_path (pc=pc@entry=0x7fffffffc870, current_format=current_format@entry=0x5555557ddd20, current_length=current_length@entry=0, max_length=3, legal_error=legal_error@entry=4.6999999999999999e-06) at ../babl-0.1.92/babl/babl-fish-path.c:749 #8 0x00007ffff7d13356 in babl_fish_path2 (source=source@entry=0x5555557ddd20, destination=destination@entry=0x555555598e40, tolerance=<optimized out>, tolerance@entry=0) at ../babl-0.1.92/babl/babl-fish-path.c:1137 #9 0x00007ffff7d13729 in babl_fish_path (source=source@entry=0x5555557ddd20, destination=destination@entry=0x555555598e40) at ../babl-0.1.92/babl/babl-fish-path.c:1209 #10 0x00007ffff7d17d46 in babl_fish (source=<optimized out>, destination=<optimized out>) at ../babl-0.1.92/babl/babl-fish.c:304 #11 0x00007ffff7e9fde1 in gegl_buffer_iterate_read_simple (buffer=0x5555557ff1c0, roi=0x7fffffffced0, buf=0x7fffda5df060 "", buf_stride=1024, format=0x555555598e40, level=0) at ../gegl-0.4.36/gegl/buffer/gegl-buffer-access.c:971 #12 0x00007ffff7ea2863 in gegl_buffer_iterate_read_dispatch (buffer=0x5555557ff1c0, roi=0x7fffffffcfe0, buf=0x7fffda5df060 "", rowstride=1024, format=0x555555598e40, level=0, repeat_mode=GEGL_ABYSS_NONE) at ../gegl-0.4.36/gegl/buffer/gegl-buffer-access.c:1864 #13 0x00007ffff7ea41d7 in gegl_buffer_copy2 (src=0x5555557ff1c0, src_rect=0x7fffffffd0d0, repeat_mode=GEGL_ABYSS_NONE, dst=0x555555806050, dst_rect=0x7fffffffd0e0) at ../gegl-0.4.36/gegl/buffer/gegl-buffer-access.c:2374 #14 0x00007ffff7ea514b in gegl_buffer_copy (src=0x5555557ff1c0, src_rect=0x7fffffffd0d0, repeat_mode=GEGL_ABYSS_NONE, dst=0x555555806050, dst_rect=0x7fffffffd0e0) at ../gegl-0.4.36/gegl/buffer/gegl-buffer-access.c:2646 #15 0x00007ffff567490a in process (operation=0x555555685290, context=0x5555557d8d40, output_prop=0x7ffff7f6ac4b "output", roi=0x5555557d8d50, level=0) at ../gegl-0.4.36/operations/core/cast-format.c:89 #16 0x00007ffff7ef980b in gegl_operation_process (operation=0x555555685290, context=0x5555557d8d40, output_pad=0x7ffff7f6ac4b "output", result=0x5555557d8d50, level=0) at ../gegl-0.4.36/gegl/operation/gegl-operation.c:176 #17 0x00007ffff7efedd3 in gegl_graph_process (path=0x5555557fe5c0, level=0) at ../gegl-0.4.36/gegl/process/gegl-graph-traversal.c:486 #18 0x00007ffff7efd910 in gegl_eval_manager_apply (self=0x5555557e08c0, roi=0x7fffffffd3d0, level=0) at ../gegl-0.4.36/gegl/process/gegl-eval-manager.c:128 #19 0x00007ffff7eda185 in gegl_node_blit_buffer (self=0x55555568e890, buffer=0x5555557ff080, roi=0x5555557b2680, level=0, abyss_policy=GEGL_ABYSS_NONE) at ../gegl-0.4.36/gegl/graph/gegl-node.c:1139 #20 0x00007ffff7eda598 in gegl_node_blit (self=0x55555568e890, scale=1, roi=0x5555557b2680, format=0x555555598e40, destination_buf=0x0, rowstride=1024, flags=GEGL_BLIT_CACHE) at ../gegl-0.4.36/gegl/graph/gegl-node.c:1215 #21 0x00007ffff7f00551 in render_rectangle (processor=0x5555557a83f0) at ../gegl-0.4.36/gegl/process/gegl-processor.c:513 #22 0x00007ffff7f00962 in gegl_processor_render (processor=0x5555557a83f0, rectangle=0x5555557a8418, progress=0x0) at ../gegl-0.4.36/gegl/process/gegl-processor.c:647 #23 0x00007ffff7f00e05 in gegl_processor_work (processor=0x5555557a83f0, progress=0x0) at ../gegl-0.4.36/gegl/process/gegl-processor.c:781 #24 0x00007ffff7edcfb1 in gegl_node_process (self=0x55555568ef10) at ../gegl-0.4.36/gegl/graph/gegl-node.c:1861 #25 0x000055555555902f in main (argc=13, argv=0x7fffffffd8b8) at ../gegl-0.4.36/bin/gegl.c:379
(In reply to Sergey Torokhov from comment #1) > Could you, please, inform is this issue only for gimp-2.10.32 or was > observed early on you system (e.g. gimp-2.10.28 or 2.10.30)? I faced the problem for the first time but both gegl and gimp where updated in this round. I will rebuild media-gfx/gimp-2.10.28-r1 with the new gegl-0.4.36-r1 and report back.
(In reply to Alexander Bezrukov from comment #3) > (In reply to Sergey Torokhov from comment #1) > > Could you, please, inform is this issue only for gimp-2.10.32 or was > > observed early on you system (e.g. gimp-2.10.28 or 2.10.30)? > > I faced the problem for the first time but both gegl and gimp where updated > in this round. > > I will rebuild media-gfx/gimp-2.10.28-r1 with the new gegl-0.4.36-r1 and > report back. It looks similar about "segfalt", "R'G'B'A float" and "opteron-sse3". Is it reproduced if disable for babl, gegl this "opteron-sse3" option. The CPU is rather old (2006 year) and it's sse could be specific. The <2.10.32 are CVEs affected and are planning to be dropped soon.
Sorry, i.e. it looks familiar (not "similar") and already seen issue.
(In reply to Sergey Torokhov from comment #5) > Sorry, i.e. it looks familiar (not "similar") and already seen issue. Is this the same issue described on one of the linux society forum? : https://www.linux.org.ru/forum/multimedia/16845219
(In reply to Alexander Bezrukov from comment #3) > I will rebuild media-gfx/gimp-2.10.28-r1 with the new gegl-0.4.36-r1 and > report back. The problem currently exists with all combinations of compatible versions of gegl and gimp. (In reply to Sergey Torokhov from comment #4) > (In reply to Alexander Bezrukov from comment #3) > It looks similar about "segfalt", "R'G'B'A float" and "opteron-sse3". > Is it reproduced if disable for babl, gegl this "opteron-sse3" option. > The CPU is rather old (2006 year) and it's sse could be specific. Yes, the name of the function universal_nonlinear_rgb_u8_converter_sse2 implies sse2 but the CPU (it is Opteron 290) supports sse2.(In reply to Sergey Torokhov from comment #6) > (In reply to Sergey Torokhov from comment #5) > Is this the same issue described on one of the linux society forum? : > https://www.linux.org.ru/forum/multimedia/16845219 There is no much details on that forum post. It looks very similar at least. One of the commenters suggests to rebuild the whole world with -march=opteron-sse3 (which is clang syntax, not that of gcc, as far as I know) and the topic starter has actually done that excercise to no avail. I will try to debug further when I can more time.
Please attach the gegl-0.4.36-r1 build.log I.e. run command `ebuild /usr/portage/media-libs/gegl/gegl-0.4.26-r1 clean compile` and atach resulted build.log file. It's very strange error and looks like affected Opteron K8 processes. According [1] for K8 CPUs there are additionaly were ‘k8-sse3’, ‘opteron-sse3’, ‘athlon64-sse3’ besides ‘k8’, ‘opteron’, ‘athlon64’, ‘athlon-fx’ options. As example I had issue to build media-libs/libaom with SSE2 but has no SSSE3 (not SSE3 but SSSE3) while my CPU support doesn't support SSSE3 [2]. Maybe the disabling any sse3 for babl, gegl and gimp could be help. It's strange that in gimp build.log > checking whether x86_64-pc-linux-gnu-gcc understands -msse4.1... yes > checking whether we can compile SSE4.1 intrinsics... yes while this Opteron 290 doesn't support it. [1] https://gcc.gnu.org/onlinedocs/gcc/x86-Options.html [2] https://bugs.gentoo.org/671340
Created attachment 791126 [details] The buld log of gegl-0.4.36-r1 (In reply to Sergey Torokhov from comment #8) > Please attach the gegl-0.4.36-r1 build.log Done. This variant is compiled with conservative -march=k8 (not -march=native, which is equivalent -march=opteron-sse3, and I was wrong: gcc supports this syntax, it was not the case years ago but now it is). This build is for debug only but I checked that the segfault manifests with this build. Yes I see -mssse3 in the build log. What also I see important is the problem manifests as segmentation fault and not as illegal instruction -- the manifestation I would expect if compiler emits an unsupported instruction. > According [1] for K8 CPUs there are additionaly were ‘k8-sse3’, > ‘opteron-sse3’, ‘athlon64-sse3’ besides ‘k8’, ‘opteron’, ‘athlon64’, > ‘athlon-fx’ options. Yes. And my CPU is *-sse3 variant, it supports sse3 (aka pni). > As example I had issue to build media-libs/libaom with SSE2 but has no SSSE3 > (not SSE3 but SSSE3) while my CPU support doesn't support SSSE3 [2]. I am aware of that issue and my version is manually patched. > It's strange that in gimp build.log > > checking whether x86_64-pc-linux-gnu-gcc understands -msse4.1... yes > > checking whether we can compile SSE4.1 intrinsics... yes It's not that strange: the compiler obviously understands -msse4.1 and it can compile any modern intrinsics because the compiler is sufficiently modern. Hopefully, that tests are used for good and gimp does not assume that these instructions exist on the CPU just because the compiler is aware of their existence in the big world. > while this Opteron 290 doesn't support it. Opteron doesn't but gcc does.
I looked slightly deeper. The actual crash in babl in matrix-vector multiplication code (MMX variant) babl_matrix_mul_vectorff_buf4_sse2 (babl-0.1.92/babl/base/babl-rgb-converter.c:244) and it is a little cryptic. gdb shows that segfault happens during fetching 128-bit vector from memory (gcc's xmmintrin.h:927) at address 0x7fffffffbff8. The address actually contains a vector ({0.77423048, 0.751898408, 0.923993587, 3.06043585e-41}) but the address is obviously not 16-byte aligned. What I would expect is bus error, not segmentation fault, at least I used to catch bus errors in similar situations in other unices. The stack when all functions and intrinsics are un-inlined is as follows: Thread 1 "gegl-0.4" received signal SIGSEGV, Segmentation fault. 0x00007ffff7d31af0 in _mm_load_ps (__P=0x7fffffffbff8) at /usr/lib/gcc/x86_64-pc-linux-gnu/11.3.0/include/xmmintrin.h:927 927 return *(__m128 *)__P; (gdb) where #0 0x00007ffff7d31af0 in _mm_load_ps (__P=0x7fffffffbff8) at /usr/lib/gcc/x86_64-pc-linux-gnu/11.3.0/include/xmmintrin.h:927 #1 babl_matrix_mul_vectorff_buf4_sse2 (mat=0x5555558db900, v_in=0x7fffffffbff8, v_out=0x7fffffffbff8, samples=32) at ../babl-0.1.92/babl/base/babl-rgb-converter.c:244 #2 0x00007ffff7d3210b in universal_nonlinear_rgb_u8_converter_sse2 (conversion=0x555555946710, src_char=0x555555943df0 "\343\340\366[\252\326\372\\\240", dst_char=0x555555943d50 "", samples=32, data=0x5555558db900) at ../babl-0.1.92/babl/base/babl-rgb-converter.c:318 #3 0x00007ffff7cfaea4 in _babl_process (cbabl=0x555555946710, source=0x555555943df0, destination=0x555555943d50, n=32) at ../babl-0.1.92/babl/babl-fish-path.c:1323 #4 0x00007ffff7cfaedf in babl_process (babl=0x555555946710, source=0x555555943df0, destination=0x555555943d50, n=32) at ../babl-0.1.92/babl/babl-fish-path.c:1333 #5 0x00007ffff7cf4959 in babl_conversion_error (conversion=0x5555555aa520) at ../babl-0.1.92/babl/babl-conversion.c:504 #6 0x00007ffff7cf966f in get_conversion_path (pc=0x7fffffffc7f0, current_format=0x555555598e40, current_length=3, max_length=3, legal_error=4.6999999999999999e-06) at ../babl-0.1.92/babl/babl-fish-path.c:690 #7 0x00007ffff7cf9904 in get_conversion_path (pc=0x7fffffffc7f0, current_format=0x5555555964f0, current_length=2, max_length=3, legal_error=4.6999999999999999e-06) at ../babl-0.1.92/babl/babl-fish-path.c:749 #8 0x00007ffff7cf9904 in get_conversion_path (pc=0x7fffffffc7f0, current_format=0x555555610c30, current_length=1, max_length=3, legal_error=4.6999999999999999e-06) at ../babl-0.1.92/babl/babl-fish-path.c:749 #9 0x00007ffff7cf9904 in get_conversion_path (pc=0x7fffffffc7f0, current_format=0x5555557d0200, current_length=0, max_length=3, legal_error=4.6999999999999999e-06) at ../babl-0.1.92/babl/babl-fish-path.c:749 #10 0x00007ffff7cfa7fd in babl_fish_path2 (source=0x5555557d0200, destination=0x555555598e40, tolerance=4.6999999999999999e-06) at ../babl-0.1.92/babl/babl-fish-path.c:1137 #11 0x00007ffff7cfab48 in babl_fish_path (source=0x5555557d0200, destination=0x555555598e40) at ../babl-0.1.92/babl/babl-fish-path.c:1209 #12 0x00007ffff7d01668 in babl_fish (source=0x5555557d0200, destination=0x555555598e40) at ../babl-0.1.92/babl/babl-fish.c:304 #13 0x00007ffff7ed9de1 in gegl_buffer_iterate_read_simple (buffer=0x5555557d31c0, roi=0x7fffffffce60, buf=0x7fffd9dde060 "", buf_stride=1024, format=0x555555598e40, level=0) at ../gegl-0.4.36/gegl/buffer/gegl-buffer-access.c:971 #14 0x00007ffff7edc863 in gegl_buffer_iterate_read_dispatch (buffer=0x5555557d31c0, roi=0x7fffffffcf70, buf=0x7fffd9dde060 "", rowstride=1024, format=0x555555598e40, level=0, repeat_mode=GEGL_ABYSS_NONE) at ../gegl-0.4.36/gegl/buffer/gegl-buffer-access.c:1864 #15 0x00007ffff7ede1d7 in gegl_buffer_copy2 (src=0x5555557d31c0, src_rect=0x7fffffffd060, repeat_mode=GEGL_ABYSS_NONE, dst=0x5555557db050, dst_rect=0x7fffffffd070) at ../gegl-0.4.36/gegl/buffer/gegl-buffer-access.c:2374 #16 0x00007ffff7edf14b in gegl_buffer_copy (src=0x5555557d31c0, src_rect=0x7fffffffd060, repeat_mode=GEGL_ABYSS_NONE, dst=0x5555557db050, dst_rect=0x7fffffffd070) at ../gegl-0.4.36/gegl/buffer/gegl-buffer-access.c:2646 #17 0x00007ffff564090a in process (operation=0x555555652e90, context=0x5555557b7190, output_prop=0x7ffff7fa6c4b "output", roi=0x5555557b71a0, level=0) at ../gegl-0.4.36/operations/core/cast-format.c:89 #18 0x00007ffff7f3380b in gegl_operation_process (operation=0x555555652e90, context=0x5555557b7190, output_pad=0x7ffff7fa6c4b "output", result=0x5555557b71a0, level=0) at ../gegl-0.4.36/gegl/operation/gegl-operation.c:176 #19 0x00007ffff7f38dd3 in gegl_graph_process (path=0x5555558044f0, level=0) at ../gegl-0.4.36/gegl/process/gegl-graph-traversal.c:486 #20 0x00007ffff7f37910 in gegl_eval_manager_apply (self=0x5555557b5cc0, roi=0x7fffffffd360, level=0) at ../gegl-0.4.36/gegl/process/gegl-eval-manager.c:128 #21 0x00007ffff7f14185 in gegl_node_blit_buffer (self=0x55555568e090, buffer=0x5555557d3080, roi=0x5555557b0b80, level=0, abyss_policy=GEGL_ABYSS_NONE) at ../gegl-0.4.36/gegl/graph/gegl-node.c:1139 #22 0x00007ffff7f14598 in gegl_node_blit (self=0x55555568e090, scale=1, roi=0x5555557b0b80, format=0x555555598e40, destination_buf=0x0, rowstride=1024, flags=GEGL_BLIT_CACHE) at ../gegl-0.4.36/gegl/graph/gegl-node.c:1215 #23 0x00007ffff7f3a551 in render_rectangle (processor=0x5555557a6bf0) at ../gegl-0.4.36/gegl/process/gegl-processor.c:513 #24 0x00007ffff7f3a962 in gegl_processor_render (processor=0x5555557a6bf0, rectangle=0x5555557a6c18, progress=0x0) at ../gegl-0.4.36/gegl/process/gegl-processor.c:647 #25 0x00007ffff7f3ae05 in gegl_processor_work (processor=0x5555557a6bf0, progress=0x0) at ../gegl-0.4.36/gegl/process/gegl-processor.c:781 #26 0x00007ffff7f16fb1 in gegl_node_process (self=0x55555568e710) at ../gegl-0.4.36/gegl/graph/gegl-node.c:1861 #27 0x000055555555902f in main (argc=13, argv=0x7fffffffd848) at ../gegl-0.4.36/bin/gegl.c:379 (gdb) p *(__m128 *)__P $4 = {0.77423048, 0.751898408, 0.923993587, 3.06043585e-41} The actual instruction is movaps (%rax), %xmm0 and %rax while seemingly points to correct data is indeed not aligned. Probably this is the culprit.
The fact that the problem is Opteron-specific can be explained by aliasing aligned and non-aligned instructions on non-server CPUs. If this explanation correct, then the same bug must manifest also on Xeons, at least if the same mmx variant of matrix-vector multiplication is used. I have couple of machines with Xeons but it would require quite an amount of work (to install Gentoo, first of all) to try to reproduce the issue.
May want to try ubsan as it can clearly mark unaligned accesses.
(In any case, please report this upstream & link it here, if you wouldn't mind)
(In reply to Alexander Bezrukov from comment #11) > The fact that the problem is Opteron-specific can be explained by aliasing > aligned and non-aligned instructions on non-server CPUs. If this explanation > correct, then the same bug must manifest also on Xeons, at least if the same > mmx variant of matrix-vector multiplication is used. I have couple of > machines with Xeons but it would require quite an amount of work (to install > Gentoo, first of all) to try to reproduce the issue. I'm also confused this part of gcc/gcc/config/i386/xmmintrin.h (line 923-927): > /* Load four SPFP values from P. The address must be 16-byte aligned. */ > extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__)) > _mm_load_ps (float const *__P) > { > return *(__m128 *)__P; > } I.e. there is "float const *__P" while vector contains "3.06043585e-41" that out of float limits. Anyway I should file a bug for Babl project, as Sam offers.
(In reply to Sergey Torokhov from comment #14) > I.e. there is "float const *__P" while vector contains "3.06043585e-41" that > out of float limits. No, not quite out of limits. That is, "3.06043585e-41" is a subnormal number. But the least signigicant bits in its mantissa form an interesting pattern: 0x5550. This can of course happen with ordinary arithmetic operations. > Anyway I should file a bug for Babl project, as Sam offers. The mere fact that unaligned access has happened in babl doesn't mean that the bug is in babl. The fact that access is on the stack added to the suspicion but still it can be gegl who misaligned the stack pointer. I will devote some time first to look a little further, otherwise chances are high that such a report will create unnecessary noise. Unfortunately, I have no much time right now but hopefully I can spend some in the night.
I encountered the exact same problem in a x86 build for media-gfx/gimp-2.10.32. The relevant build log fragement is below along with a gegl stack traceback showing the gegl segfault occurred in libbabl. I looked through the issues (opened and closed) for both gimp, gegl, and babl for the past year and didn't see anything related to this issue. However, I see in the gentoo forums that someone had a similar issue with an amd64 gimp-2.10.30 build, but they did not mention which version of gegl or babl they were using: https://forums.gentoo.org/viewtopic-t-1148735.html My builds for media-gfx/gimp-2.10.30 on both x86 and amd64 had no issues. Both of those builds used media-libs/gegl-0.4.34 and media-libs/babl-0.1.88. The x86 build that failed was using media-libs/gegl-0.4.36-r1 and media-libs/babl-0.1.92. I did not try to build media-gfx/gimp-2.10.32 on amd64 yet. ** Relevant Build Log Fragement: mkdir -p `dirname 64/gimp-texture.png` && GEGL_USE_OPENCL=no GEGL_SWAP=ram /usr/bin/gegl-0.4 ../../icons/Symbolic/64/gimp-texture.png -o 64/gimp-texture.png -- cast-format input-format="R'G'B'A float" output-format="RGBA float" brightness-contrast contrast=1.5 cast-format input-format="RGBA float" output-format="R'G'B'A float" mkdir -p `dirname 64/gimp-user-manual.png` && GEGL_USE_OPENCL=no GEGL_SWAP=ram /usr/bin/gegl-0.4 ../../icons/Symbolic/64/gimp-user-manual.png -o 64/gimp-user-manual.png -- cast-format input-format="R'G'B'A float" output-format="RGBA float" brightness-contrast contrast=1.5 cast-format input-format="RGBA float" output-format="R'G'B'A float" mkdir -p `dirname 64/gimp-wilber-eek.png` && GEGL_USE_OPENCL=no GEGL_SWAP=ram /usr/bin/gegl-0.4 ../../icons/Symbolic/64/gimp-wilber-eek.png -o 64/gimp-wilber-eek.png -- cast-format input-format="R'G'B'A float" output-format="RGBA float" brightness-contrast contrast=1.5 cast-format input-format="RGBA float" output-format="R'G'B'A float" mkdir -p `dirname 64/gimp-wilber.png` && GEGL_USE_OPENCL=no GEGL_SWAP=ram /usr/bin/gegl-0.4 ../../icons/Symbolic/64/gimp-wilber.png -o 64/gimp-wilber.png -- cast-format input-format="R'G'B'A float" output-format="RGBA float" brightness-contrast contrast=1.5 cast-format input-format="RGBA float" output-format="R'G'B'A float" make[3]: *** [Makefile:2445: 64/gimp-texture.png] Segmentation fault make[3]: *** Waiting for unfinished jobs.... make[3]: Leaving directory '/var/tmp/portage/media-gfx/gimp-2.10.32/work/gimp-2.10.32/icons/Symbolic-High-Contrast' make[2]: *** [Makefile:670: all-recursive] Error 1 make[2]: Leaving directory '/var/tmp/portage/media-gfx/gimp-2.10.32/work/gimp-2.10.32/icons' make[1]: *** [Makefile:861: all-recursive] Error 1 make[1]: Leaving directory '/var/tmp/portage/media-gfx/gimp-2.10.32/work/gimp-2.10.32' make: *** [Makefile:762: all] Error 2 ** Stack Traceback from gdb: $ gdb -q /usr/bin/gegl-0.4 core Reading symbols from /usr/bin/gegl-0.4... (No debugging symbols found in /usr/bin/gegl-0.4) [New LWP 7826] [New LWP 7827] [New LWP 7833] [Thread debugging using libthread_db enabled] Using host libthread_db library "/lib/libthread_db.so.1". Core was generated by `/usr/bin/gegl-0.4 /var/tmp/portage/media-gfx/gimp-2.10.32/work/gimp-2.10.32/ico'. Program terminated with signal SIGSEGV, Segmentation fault. #0 0xb7d93ff0 in ?? () from /usr/lib/libbabl-0.1.so.0 [Current thread is 1 (Thread 0xb74fc2c0 (LWP 7826))] (gdb) bt #0 0xb7d93ff0 in () at /usr/lib/libbabl-0.1.so.0 #1 0xb7d651f2 in babl_process () at /usr/lib/libbabl-0.1.so.0 #2 0xb7d6021d in () at /usr/lib/libbabl-0.1.so.0 #3 0xb7d63ad5 in () at /usr/lib/libbabl-0.1.so.0 #4 0xb7d63a7a in () at /usr/lib/libbabl-0.1.so.0 #5 0xb7d63a7a in () at /usr/lib/libbabl-0.1.so.0 #6 0xb7d63a7a in () at /usr/lib/libbabl-0.1.so.0 #7 0xb7d64d55 in () at /usr/lib/libbabl-0.1.so.0 #8 0xb7d6a77a in babl_fish () at /usr/lib/libbabl-0.1.so.0 #9 0xb7e99e88 in () at /usr/lib/libgegl-0.4.so.0 #10 0xb7e9c00c in () at /usr/lib/libgegl-0.4.so.0 #11 0xb7e9e5c6 in () at /usr/lib/libgegl-0.4.so.0 #12 0xb7e9fbc5 in gegl_buffer_copy () at /usr/lib/libgegl-0.4.so.0 #13 0xb6ade869 in () at /usr/lib/gegl-0.4/gegl-core.so #14 0xb7edfa47 in gegl_operation_process () at /usr/lib/libgegl-0.4.so.0 #15 0xb7ee209e in gegl_graph_process () at /usr/lib/libgegl-0.4.so.0 #16 0xb7ee0f5f in gegl_eval_manager_apply () at /usr/lib/libgegl-0.4.so.0 #17 0xb7ec59cb in gegl_node_blit_buffer () at /usr/lib/libgegl-0.4.so.0 #18 0xb7ec61f3 in gegl_node_blit () at /usr/lib/libgegl-0.4.so.0 #19 0xb7ee3670 in gegl_processor_work () at /usr/lib/libgegl-0.4.so.0 #20 0xb7ec5c2e in gegl_node_process () at /usr/lib/libgegl-0.4.so.0 #21 0x0047f124 in main () (gdb) quit
I debugged and fixed the issue. 1. Unaligned accesses are actually produce SIGSEGV instead of SIGBUS in modern kernels, this itself may be a bug in the kernel. A simple test demonstates this behavior: typedef float v4sf __attribute__((vector_size(4 * sizeof(float)))); int main() { static v4sf v[400]; v4sf *const w = (v4sf *)((float *)(v)+2); w[0]=w[1]+w[2]; return 0; } 2. The code in babl is indeed careless about alignment. I will prepare a fix for this bug and post here a patch.
Created attachment 791267 [details, diff] A trivial fix for alignment bug Sergey, I investigated and fixed the issue, would you please report the bug and the fix upstream? Would you consider applying it meanwhile in Gentoo?
(In reply to Alexander Bezrukov from comment #18) > Created attachment 791267 [details, diff] [details, diff] > A trivial fix for alignment bug was affected too locally applied proposed patch to =media-libs/babl-0.1.92 and gimp's build has eventually succeeded. * AMD Athlon(tm) Dual Core Processor 5050e, CPU family: 15, Model: 107 * =sys-devel/gcc-11.3.0 * CFLAGS="-O2 -march=k8-sse3 …"
(In reply to Alexander Bezrukov from comment #18) > Created attachment 791267 [details, diff] [details, diff] > A trivial fix for alignment bug > > Sergey, I investigated and fixed the issue, would you please report the bug > and the fix upstream? Would you consider applying it meanwhile in Gentoo? Great thanks for investigation and fix! It's seems that issue is for K8 cpu, I wasn't affected it on my K10 cpu. I will do quick tests this evening on my PC (non-K8 cpu) and will create pull request for ::gentoo portage tree. Initially the revision bump (babl-0.1.92-r1) will be non-stable. Filing a bug and pull request to upstream will take additional time for me. P.S. Thanks to all who did additional testing on K8 CPUs.
The patch also fixes the build issue on my x86 host (Intel cpu, -march=i686).
The bug has been closed via the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=c46124b30e8e1c5d9ce6a9f017c024116c34c5af commit c46124b30e8e1c5d9ce6a9f017c024116c34c5af Author: Sergey Torokhov <torokhov-s-a@yandex.ru> AuthorDate: 2022-07-13 19:36:56 +0000 Commit: Sam James <sam@gentoo.org> CommitDate: 2022-07-14 02:45:54 +0000 media-libs/babl: 0.1.92-r1. Fix universal_nonlinear_rgb_u8_converter_sse2 Thanks-to: Alexander Bezrukov <phmagic@mail.ru> Closes: https://bugs.gentoo.org/857708 Signed-off-by: Sergey Torokhov <torokhov-s-a@yandex.ru> Closes: https://github.com/gentoo/gentoo/pull/26393 Signed-off-by: Sam James <sam@gentoo.org> media-libs/babl/babl-0.1.92-r1.ebuild | 68 ++++++++++++++++++++++ ...universal_nonlinear_rgb_u8_converter_sse2.patch | 17 ++++++ 2 files changed, 85 insertions(+)
(In reply to Alexander Bezrukov from comment #18) > Created attachment 791267 [details, diff] [details, diff] > A trivial fix for alignment bug > > Sergey, I investigated and fixed the issue, would you please report the bug > and the fix upstream? Would you consider applying it meanwhile in Gentoo? The offered patch was merged into upstream: https://gitlab.gnome.org/GNOME/babl/-/commit/8c993497f338548bdb8247f4796545e250b5e05c It seems the issue is realy common for K8 CPUs as reported in several recent issues. Great thanks again!
The bug has been referenced in the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=002591c6e95035b57426ff89de229afb7e0db474 commit 002591c6e95035b57426ff89de229afb7e0db474 Author: Sergey Torokhov <torokhov-s-a@yandex.ru> AuthorDate: 2022-07-15 22:43:40 +0000 Commit: Joonas Niilola <juippis@gentoo.org> CommitDate: 2022-07-25 06:04:39 +0000 media-libs/babl: add upstream bug link to patch Add upstream bug link to 'universal_nonlinear_rgb_u8_converter_sse2.patch' Bug: https://bugs.gentoo.org/857708 Signed-off-by: Sergey Torokhov <torokhov-s-a@yandex.ru> Closes: https://github.com/gentoo/gentoo/pull/26436 Signed-off-by: Joonas Niilola <juippis@gentoo.org> .../files/babl-0.1.92_universal_nonlinear_rgb_u8_converter_sse2.patch | 1 + 1 file changed, 1 insertion(+)
*** Bug 860750 has been marked as a duplicate of this bug. ***
