LO segfaults in one test on armhf when built with GCC 10

[Iain Lane]

I don't think we've had a successful build of Libreoffce on armhf in Ubuntu since gcc-10, it always eventually segfaults after 24 hours or so

/<<PKGBUILDDIR>>/toolkit/source/controls/tkscrollbar.cxx: In member function ‘virtual void toolkit::UnoScrollBarControl::adjustmentValueChanged(const com::sun::star::awt::AdjustmentEvent&)’:
/<<PKGBUILDDIR>>/toolkit/source/controls/tkscrollbar.cxx:175:5: internal compiler error: Segmentation fault
  175 | }
      | ^
Please submit a full bug report,

[Iain Lane (laney)]

I'm running a build now to get the preprocessed stuff.

[Matthias Klose (doko)]

successfully built in the test rebuild:
https://launchpad.net/ubuntu/+archive/test-rebuild-20200728-groovy/+sourcepub/11459791/+listing-archive-extra

[Iain Lane (laney)]

Now it builds, but the resulting binaries segfault

  https://launchpad.net/~ci-train-ppa-service/+archive/ubuntu/4212/+packages

but when built using gcc-9 on armhf it works

  https://launchpad.net/~libreoffice/+archive/ubuntu/libreoffice-prereleases/+build/19834431

diff between these two:

  https://launchpadlibrarian.net/493615789/libreoffice_1%3A7.0.1~rc1-0ubuntu0.20.10.1_1%3A7.0.1~rc1-0ubuntu0.20.10.2.diff.gz

[Heather Ellsworth (hellsworth)]

/<<PKGBUILDDIR>>/dbaccess/source/core/api/StaticSet.cxx: In member function ‘virtual
com::sun::star::uno::Any dbaccess::OStaticSet::getBookmark()’:
/<<PKGBUILDDIR>>/dbaccess/source/core/api/StaticSet.cxx:47:1: internal compiler error: Segmentation fault

[Balint Reczey (rbalint)]

I've also seen occasional segfaults in other builds, glibc maybe? @laney Do those binaries segfault on local systems or on the test infrastructure?

[Iain Lane (laney)]

Hey Balint,

I didn't manage to get any binaries out so can't tell you directly. Heather took this task over and may have gotten further; I will ask later on for an update on that.

I think I did notice that sometimes the *build* succeeded but then the *tests* were crashing so maybe that does confirm what you're asking.

btw, for Libreoffice in groovy we've switched the armhf build to gcc-9 and that works fine.

[Heather Ellsworth (hellsworth)]

In my investigation, I've found that a local armhf build fails but I do not see the segmentation fault. To build, I'm just doing `dpkg-buildpackage -us -uc`. I have the build log but it's 7.5Mb of uninteresting stuff.. no smoking gun.

[Matthias Klose (doko)]

I am opening a libreoffice task, and closing the gcc-10 task. I didn't see any ICE during my debugging work. Unfortunately the package doesn't build with other optimization options due to packaging and upstream issues. Separate issues are filed as LP: #1894233, LP: #1894224, LP: #1894223, LP: #1894222.

[Rico Tzschichholz (ricotz)]

Still happening on impish with gcc-10 10.3.0-4ubuntu1 and libreoffice 7.1.4

https://launchpadlibrarian.net/543439072/buildlog_ubuntu-impish-armhf.libreoffice_1%3A7.1.4~rc2-0ubuntu1+lo3_BUILDING.txt.gz

[Matthias Klose (doko)]

any progress on the issues mentioned in #8? could you do a debug build once these are fixed?

[Brian Murray (brian-murray)]

The Groovy Gorilla has reached end of life, so this bug will not be fixed for that release

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[Brian Murray (brian-murray)]

The Hirsute Hippo has reached End of Life, so this bug will not be fixed for that release.

[Brian Murray (brian-murray)]

Ubuntu 21.10 (Impish Indri) has reached end of life, so this bug will not be fixed for that specific release.

[Launchpad Janitor (janitor)]

[Expired for libreoffice (Ubuntu) because there has been no activity for 60 days.]