Bug#930012: gcc-8: ICE building firefox 68.0~b6-2 on s390x
On Wed, Jun 05, 2019 at 06:26:42PM +0900, Mike Hommey wrote: > Control: tags -1 -moreinfo > > On Wed, Jun 05, 2019 at 10:49:16AM +0200, Matthias Klose wrote: > > Control: tags -1 moreinfo > > > > it's always appreciated to attach the preprocessed source, together with the > > command line options used. > > Attached is the full cc*.out file output by a build I did locally (maybe > buildds should upload those automatically somewhere). > > > > please also check with gcc-8 and gcc-9 from experimental. > > Both ICE the same way. This was fixed upstream. gcc-9 needs PR90756 to work, and older versions both PR90756 and PR90139. Any chance we can apply them to the packages in unstable? What about now-stable buster? Mike
Bug#930012: gcc-8: ICE building firefox 68.0~b6-2 on s390x and i386
Workaround submitted upstream: https://bugs.chromium.org/p/skia/issues/detail?id=9202
Bug#930012: gcc-8: ICE building firefox 68.0~b6-2 on s390x and i386
On Tue, Jun 25, 2019 at 10:46:37AM +0900, Mike Hommey wrote: > On Thu, Jun 13, 2019 at 05:03:33PM +0200, Olivier Tilloy wrote: > > Ubuntu is also affected (all supported releases, from 16.04 to 19.10, > > builds are being done in > > https://launchpad.net/~mozillateam/+archive/ubuntu/firefox-next/+packages). > > > > I don't have a workaround, but here are some data points: > > - builds on 16.04 exhibit the same problem, this is with GCC 6.4.0 > > - https://skia.googlesource.com/skcms/+/7362d88 is the change in skcms (a > > skia subproject) that triggers the build error, it is the commit > > immediately after the version that's in firefox 67 source tarballs > > FWIW, reverting that change only doesn't work around the issue. Sorry, forget about this, my test was doing the opposite: apply that change to 67, and that did break the build. Mike
Bug#930012: gcc-8: ICE building firefox 68.0~b6-2 on s390x and i386
On Thu, Jun 13, 2019 at 05:03:33PM +0200, Olivier Tilloy wrote: > Ubuntu is also affected (all supported releases, from 16.04 to 19.10, > builds are being done in > https://launchpad.net/~mozillateam/+archive/ubuntu/firefox-next/+packages). > > I don't have a workaround, but here are some data points: > - builds on 16.04 exhibit the same problem, this is with GCC 6.4.0 > - https://skia.googlesource.com/skcms/+/7362d88 is the change in skcms (a > skia subproject) that triggers the build error, it is the commit > immediately after the version that's in firefox 67 source tarballs FWIW, reverting that change only doesn't work around the issue. Mike
Bug#930012: gcc-8: ICE building firefox 68.0~b6-2 on s390x and i386
On Thu, Jun 20, 2019 at 07:17:26AM +0200, Olivier Tilloy wrote: > I am attaching a patch for skcms that fixes the firefox build on s390x > and i386. Not submitted upstream yet. Note that mips and mipsel are affected as well. Mike
Bug#930012: gcc-8: ICE building firefox 68.0~b6-2 on s390x and i386
I am attaching a patch for skcms that fixes the firefox build on s390x and i386. Not submitted upstream yet. Description: work around a GCC bug that results in build failures on i386 and s390x Author: Olivier Tilloy Bug: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90756 Bug-Debian: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=930012 --- a/gfx/skia/skia/third_party/skcms/src/Transform_inl.h +++ b/gfx/skia/skia/third_party/skcms/src/Transform_inl.h @@ -559,7 +559,7 @@ SI void sample_clut_16(const skcms_A2B* // GCC 7.2.0 hits an internal compiler error with -finline-functions (or -O3) // when targeting MIPS 64, I think attempting to inline clut() into exec_ops(). -#if 1 && defined(__GNUC__) && !defined(__clang__) && defined(__mips64) +#if 1 && defined(__GNUC__) && !defined(__clang__) && (defined(__mips64) || defined(__i386) || defined(__s390x__)) #define MAYBE_NOINLINE __attribute__((noinline)) #else #define MAYBE_NOINLINE
Bug#930012: gcc-8: ICE building firefox 68.0~b6-2 on s390x and i386
Ubuntu is also affected (all supported releases, from 16.04 to 19.10, builds are being done in https://launchpad.net/~mozillateam/+archive/ubuntu/firefox-next/+packages). I don't have a workaround, but here are some data points: - builds on 16.04 exhibit the same problem, this is with GCC 6.4.0 - https://skia.googlesource.com/skcms/+/7362d88 is the change in skcms (a skia subproject) that triggers the build error, it is the commit immediately after the version that's in firefox 67 source tarballs - reverting skcms to the version in the 67 source tarballs (revision dfd5b3a) makes the build succeed, with no apparent functional issues (tested on i386) - for reference, this is the range of new skcms revisions in the firefox 68 source tarballs: https://skia.googlesource.com/skcms/+log/dfd5b3a..5e67e5c
Bug#930012: gcc-8: ICE building firefox 68.0~b6-2 on s390x
Control: tags -1 -moreinfo On Wed, Jun 05, 2019 at 10:49:16AM +0200, Matthias Klose wrote: > Control: tags -1 moreinfo > > it's always appreciated to attach the preprocessed source, together with the > command line options used. Attached is the full cc*.out file output by a build I did locally (maybe buildds should upload those automatically somewhere). > > please also check with gcc-8 and gcc-9 from experimental. Both ICE the same way. Mike cc2crbqx.out.gz Description: application/gzip
Bug#930012: gcc-8: ICE building firefox 68.0~b6-2 on s390x
Control: tags -1 moreinfo it's always appreciated to attach the preprocessed source, together with the command line options used. please also check with gcc-8 and gcc-9 from experimental. On 05.06.19 07:42, Mike Hommey wrote: > On Wed, Jun 05, 2019 at 02:32:10PM +0900, Mike Hommey wrote: >> Package: gcc-8 >> Version: 8.3.0-7 >> Severity: normal >> >> See >> https://buildd.debian.org/status/fetch.php?pkg=firefox&arch=s390x&ver=68.0%7Eb6-2&stamp=1559711675&raw=0 >> >> /<>/gfx/skia/skia/third_party/skcms/src/Transform_inl.h: In >> function 'void baseline::exec_ops(const Op*, const void**, const char*, >> char*, int)': >> /<>/gfx/skia/skia/third_party/skcms/src/Transform_inl.h:640:13: >> internal compiler error: in convert_move, at expr.c:218 >> static void exec_ops(const Op* ops, const void** args, >> ^~~~ >> Please submit a full bug report, >> with preprocessed source if appropriate. >> See for instructions. >> Preprocessed source stored into /tmp/ccqWUX5y.out file, please attach this >> to your bugreport. > > It also happens on i386: > https://buildd.debian.org/status/fetch.php?pkg=firefox&arch=i386&ver=68.0%7Eb6-2&stamp=1559712011&raw=0 > > Mike >
Bug#930012: gcc-8: ICE building firefox 68.0~b6-2 on s390x
On Wed, Jun 05, 2019 at 02:32:10PM +0900, Mike Hommey wrote: > Package: gcc-8 > Version: 8.3.0-7 > Severity: normal > > See > https://buildd.debian.org/status/fetch.php?pkg=firefox&arch=s390x&ver=68.0%7Eb6-2&stamp=1559711675&raw=0 > > /<>/gfx/skia/skia/third_party/skcms/src/Transform_inl.h: In > function 'void baseline::exec_ops(const Op*, const void**, const char*, > char*, int)': > /<>/gfx/skia/skia/third_party/skcms/src/Transform_inl.h:640:13: > internal compiler error: in convert_move, at expr.c:218 > static void exec_ops(const Op* ops, const void** args, > ^~~~ > Please submit a full bug report, > with preprocessed source if appropriate. > See for instructions. > Preprocessed source stored into /tmp/ccqWUX5y.out file, please attach this to > your bugreport. It also happens on i386: https://buildd.debian.org/status/fetch.php?pkg=firefox&arch=i386&ver=68.0%7Eb6-2&stamp=1559712011&raw=0 Mike
Bug#930012: gcc-8: ICE building firefox 68.0~b6-2 on s390x
Package: gcc-8 Version: 8.3.0-7 Severity: normal See https://buildd.debian.org/status/fetch.php?pkg=firefox&arch=s390x&ver=68.0%7Eb6-2&stamp=1559711675&raw=0 /<>/gfx/skia/skia/third_party/skcms/src/Transform_inl.h: In function 'void baseline::exec_ops(const Op*, const void**, const char*, char*, int)': /<>/gfx/skia/skia/third_party/skcms/src/Transform_inl.h:640:13: internal compiler error: in convert_move, at expr.c:218 static void exec_ops(const Op* ops, const void** args, ^~~~ Please submit a full bug report, with preprocessed source if appropriate. See for instructions. Preprocessed source stored into /tmp/ccqWUX5y.out file, please attach this to your bugreport.