[PATCH] test: Fix and stabilize tst-wcsmbs-clone-overflow test

[Frederic Berat]

On Tue, May 26, 2026 at 8:39 PM Adhemerval Zanella Netto <
adhemerval.zanella@linaro.org> wrote:

>
>
> On 26/05/26 14:35, Carlos O'Donell wrote:
> > On 5/26/26 8:54 AM, Adhemerval Zanella Netto wrote:
> >>
> >>
> >> On 26/05/26 08:45, Frédéric Bérat wrote:
> >>> The test tst-wcsmbs-clone-overflow was initially added to tests-static.
> >>> However, this causes the test to be unstable because gconv modules
> >>> dynamically load libc.so. Any discrepancy between the statically linked
> >>> version and the dynamically loaded one can lead to a crash.
> >>
> >> It is not clear when this would happen, is it the case that the tests
> >> is loading the system gconv modules instead of the one in the build
> >> directory?
> >
> > No.
> >
> > The wide character FILE* needs a gconv converter to convert the incoming
> > multi-byte character string, and so it needs to call dlopen, but for all
> > static binaries this requires loading the system libc.so.6.
> >
> > DJ's suggestion is that we wrap all tests-static execution with a
> container
> > in order to make sure they never load the system libraries. That would
> > certainly be a good belt-and-suspenders for testing. However, it would
> > limit testing in situations where user-namespace containers are disabled
> > (like more hardened build systems).
> >
>
> Right, I forgot that it uses the PT_INTERP. In this case, moving to a
> dynamic linked test does work.  And I agree that tests-static that
> might call dlopen should be ran in a test-container.
>

Do you still have any concerns, or can I apply the patch?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://sourceware.org/pipermail/libc-alpha/attachments/20260528/beba1a56/attachment.htm>