Re: [PATCH v3 0/5] fcntl fortification

[Frederic Berat <fberat@redhat.com>]

On Tue, Jun 20, 2023 at 2:50 PM Adhemerval Zanella Netto
<adhemerval.zanella@linaro.org> wrote:
>
>
>
> On 20/06/23 02:59, Maxim Kuvyrkov wrote:
> >> On Jun 19, 2023, at 22:36, Adhemerval Zanella Netto <adhemerval.zanella@linaro.org> wrote:
> >>
> >>
> >>
> >> On 19/06/23 11:23, Sergey Bugaev via Libc-alpha wrote:
> >>> On Mon, Jun 19, 2023 at 3:58 PM Carlos O'Donell <carlos@redhat.com> wrote:
> >>>> Pre-commit CI regression caused by this series on both aarch64 and aarch32.
> >>>> https://patchwork.sourceware.org/project/glibc/patch/20230617222218.642125-6-bugaevc@gmail.com/
> >>>>
> >>>> # Running glibc:io ...
> >>>> # FAIL: io/check-installed-headers-c
> >>>> # FAIL: io/check-installed-headers-cxx
> >>>> #
> >>>> # Running glibc:misc ...
> >>>> # FAIL: misc/check-installed-headers-c
> >>>> # FAIL: misc/check-installed-headers-cxx
> >>>> #
> >>>> # Running glibc:rt ...
> >>>> # FAIL: rt/check-installed-headers-c
> >>>> # FAIL: rt/check-installed-headers-cxx
> >>>
> >>> Thank you :(
> >>>
> >>> I'm a little bit lost in the Jenkins web UI -- is there anywhere I can
> >>> view the output of the failing tests to understand what exactly went
> >>> wrong? I see the sum files here [0], but not the .out files.
> >>>
> >>> [0]: https://ci.linaro.org/job/tcwg_glibc_check--master-arm-build/386/artifact/artifacts/artifacts.precommit/sumfiles/
> >
> > Hi Sergey,
> >
> > See the attached tarball with output from failed tests:
> > ===
> > Running glibc:io ...
> > FAIL: io/check-installed-headers-c
> > FAIL: io/check-installed-headers-cxx
> >
> > Running glibc:misc ...
> > FAIL: misc/check-installed-headers-c
> > FAIL: misc/check-installed-headers-cxx
> >
> > Running glibc:rt ...
> > FAIL: rt/check-installed-headers-c
> > FAIL: rt/check-installed-headers-cxx
> > ===
> >
> > Let me know if you need any help in reproducing these.  The results are for Ubuntu 22.04 armhf target, but results for Ubuntu 22.04 aarch64 show the same regressions.
> >
> > Interestingly the top-level check-installed-headers-* tests PASS:
> > ===
> > PASS: check-installed-headers-c
> > PASS: check-installed-headers-cxx
> > ===
> >
> >>
> >> I could not reproduce it locally, so I think it is a limitation of our
> >> (Linaro) CI framework, where not all tests artifacts are being removed
> >> with make test-clean. Maxim already sent a fix for this [1].
> >>
> >> [1] https://patchwork.sourceware.org/project/glibc/patch/20230615152547.2924770-1-maxim.kuvyrkov@linaro.org/
> >
> > Hi Adhemerval,
> >
> > We (Linaro) have fixed or added workarounds for all known testsuite stability problems in our CI before enabling reporting for failed patches.  I am surprised that you could not reproduce the above failures in your setup; let's troubleshoot your local setup offline.
>
> I think I figure out what is happening here. I am using toolchains built with
> scripts/build-many-glibcs.py which does not set fortify as default, different
> than distro that are now setting the it as default.
>
> Since in our CI we use the system compiler to check for header, fortify will
> be always enabled and thus breaking the testing.
>
> Frédéric is fixing a lot of tests to assume fortify is enabled as default, so
> maybe it is something that he is already tracking.  In any case, I think we
> should fix scripts/check-installed-headers.sh to check for different levels
> of _FORTIFY_SOURCE along with already in places flags.
>

I'm indeed working on a patch series to enable _FORTIFY_SOURCE while
building Glibc, which may improve your situation.
I expect the patchset will be out in the next few days (or even hours
if everything goes right), so you might be able to check soon.