** Description changed: - I'm still gathering information about this bug. + DFLTCC on s390x is a feature that was disabled to address an FTBFS in + that code. This was done before feature freeze in [1]. - UPDATE #1: see comment #2, this started out as an FTBFS. - UPDATE #2: filed upstream bug at https://debbugs.gnu.org/cgi/bugreport.cgi?bug=76644 + We later found out the real fix[2] for the FTBFS, and it's in an MP attached to this bug. The diff is two fold: + - a revert of [1] + - this change: + --- a/debian/rules + +++ b/debian/rules + @@ -31,8 +31,12 @@ ifeq ($(buildarch), amd64) + DEB_CPPFLAGS_MAINT_APPEND := -DUNALIGNED_OK + endif + + +# For why -Dalignas, see: + +# https://debbugs.gnu.org/cgi/bugreport.cgi?bug=66709#10 + +# and + +# LP: #2100598 + ifeq (${DEB_HOST_ARCH},s390x) + -DEB_CPPFLAGS_MAINT_APPEND := -DDFLTCC -DDFLTCC_LEVEL_MASK=0x7e + +DEB_CPPFLAGS_MAINT_APPEND := -DDFLTCC -DDFLTCC_LEVEL_MASK=0x7e -Dalignas=_Alignas + endif + + # Automatic dpkg-buildflags selection, backport-friendly - Reportedly there was a SIGILL on gzip/s390x such that the hardware - acceleration on that platform for gzip had to be disabled in an - upload[1]. - I found a matrix thread[2] about it. + PPA[3] with that fix, showing an s390x and amd64 builds. + DEP8 tests will be run after the PPA publishes, but they are rather trivial. - Looks like it happened during a PPA build[3]: - (...) - /usr/bin/make check-recursive - make[2]: Entering directory '/<<PKGBUILDDIR>>/builddir' - Making check in lib - make[3]: Entering directory '/<<PKGBUILDDIR>>/builddir/lib' - /usr/bin/make check-am - make[4]: Entering directory '/<<PKGBUILDDIR>>/builddir/lib' - make[4]: Leaving directory '/<<PKGBUILDDIR>>/builddir/lib' - make[3]: Leaving directory '/<<PKGBUILDDIR>>/builddir/lib' - Making check in doc - make[3]: Entering directory '/<<PKGBUILDDIR>>/builddir/doc' - make[3]: Nothing to be done for 'check'. - make[3]: Leaving directory '/<<PKGBUILDDIR>>/builddir/doc' - Making check in . - make[3]: Entering directory '/<<PKGBUILDDIR>>/builddir' - /usr/bin/make check-local - make[4]: Entering directory '/<<PKGBUILDDIR>>/builddir' - ./gzip < ../gzip.doc >gzip.doc.gz-t && mv gzip.doc.gz-t gzip.doc.gz - PATH=.:$PATH; { test '..' != . \ - || zdiff gzip.doc.gz; } - PATH=.:$PATH; zdiff ../gzip.doc ../gzip.doc - PATH=.:$PATH; zdiff ../gzip.doc gzip.doc.gz - Illegal instruction (core dumped) - (...) + This feature was disabled by mistake, one could say, and is now being re-enabled. Still, it's a change in behavior post FF, and warrants an FFe request. - Apparently it hapepns[4] when "#define <stdalign.h>" is added? Perhaps - to fix an FTBFS in plucky? - - """ - if you want to reproduce, take the upload Simon just sponsored, re-enable the DFLTCC instruction in debian/rules, then add #define <stdalign.h> to the includes at the top of dfltcc.c as a Quilt patch so that the code actually builds, then build it in a PPA with s390x enabled, and that'll happen. Probably you can unearth the ~danger2 build from my PPA though - """ 1. https://launchpad.net/ubuntu/+source/gzip/1.13-1ubuntu2 - 2. https://matrix.to/#/!QMtJBibTYYOCvXJEdv:ubuntu.com/$XXeSHgSv1gpBMQ2fG_8S9ZJWjP31rsXY8qY2cCmFkrI?via=ubuntu.com&via=matrix.org&via=matrix.debian.social - 3. https://launchpadlibrarian.net/776055235/buildlog_ubuntu-plucky-s390x.gzip_1.13-1ubuntu2~danger2_BUILDING.txt.gz - 4. https://matrix.to/#/!QMtJBibTYYOCvXJEdv:ubuntu.com/$QuKmfVGi7rtB8LXGC5fkiBDsg2ydqABLB3eWEtthgVw?via=ubuntu.com&via=matrix.org&via=matrix.debian.social + 2. https://debbugs.gnu.org/cgi/bugreport.cgi?bug=66709#10 + 3. https://launchpad.net/~ahasenack/+archive/ubuntu/gzip-sigill/
-- You received this bug notification because you are a member of Ubuntu Touch seeded packages, which is subscribed to gzip in Ubuntu. https://bugs.launchpad.net/bugs/2100598 Title: [FFe]: re-enable DFLTCC on s390x Status in Ubuntu on IBM z Systems: New Status in gzip package in Ubuntu: In Progress Status in gzip source package in Plucky: In Progress Bug description: DFLTCC on s390x is a feature that was disabled to address an FTBFS in that code. This was done before feature freeze in [1]. We later found out the real fix[2] for the FTBFS, and it's in an MP attached to this bug. The diff is two fold: - a revert of [1] - this change: --- a/debian/rules +++ b/debian/rules @@ -31,8 +31,12 @@ ifeq ($(buildarch), amd64) DEB_CPPFLAGS_MAINT_APPEND := -DUNALIGNED_OK endif +# For why -Dalignas, see: +# https://debbugs.gnu.org/cgi/bugreport.cgi?bug=66709#10 +# and +# LP: #2100598 ifeq (${DEB_HOST_ARCH},s390x) -DEB_CPPFLAGS_MAINT_APPEND := -DDFLTCC -DDFLTCC_LEVEL_MASK=0x7e +DEB_CPPFLAGS_MAINT_APPEND := -DDFLTCC -DDFLTCC_LEVEL_MASK=0x7e -Dalignas=_Alignas endif # Automatic dpkg-buildflags selection, backport-friendly PPA[3] with that fix, showing an s390x and amd64 builds. DEP8 tests will be run after the PPA publishes, but they are rather trivial. This feature was disabled by mistake, one could say, and is now being re-enabled. Still, it's a change in behavior post FF, and warrants an FFe request. 1. https://launchpad.net/ubuntu/+source/gzip/1.13-1ubuntu2 2. https://debbugs.gnu.org/cgi/bugreport.cgi?bug=66709#10 3. https://launchpad.net/~ahasenack/+archive/ubuntu/gzip-sigill/ To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu-z-systems/+bug/2100598/+subscriptions -- Mailing list: https://launchpad.net/~touch-packages Post to : touch-packages@lists.launchpad.net Unsubscribe : https://launchpad.net/~touch-packages More help : https://help.launchpad.net/ListHelp
