[PATCH] pthread_detach: add a comment about a known segfault bug.

[Sam James]

Carlos O'Donell <codonell@redhat.com> writes:

> On 11/24/25 5:25 PM, Aaron Jacobs wrote:
>> On Tue, Nov 25, 2025 at 1:06 AM Adhemerval Zanella Netto
>> <adhemerval.zanella@linaro.org> wrote:
>>> And I do not think adding a comment to bug report improves much here. We
>>> already have a bug report with a lot of background, and a proposed solution.
>> Right, but the bug report is not discoverable from the source code,
>> which
>> somebody like me may spend hours looking at before luckily finding the bug
>> report from some random Reddit thread when they finally get the right search
>> term. This is not a good experience.
>
> The 4th hit in google for "pthread_detach known segfault" is bug 19951.
>
>> This also happened to the other very experienced engineer with lots of
>> debugging experience in glibc and elsewhere who was assisting me.
>> What is the harm of making it more discoverable? This can easily be
>> reverted
>> with the fix for the bug.
>
> There is no objective harm in making it more discoverable.
>
> We have in the past documented known issues in the sources with links to bugs.
>
> The point Adhmerveral makes is that we could just review the fix?
>
> It's a matter of taste here, to put effort into the fix.
>
> If we get close to the release without a fix, then we should put in some comments.

+1 on this compromise, just don't want to really take momentum out of us
testing and reviewing azanella's fix instead.

>
>  > On Tue, Nov 25, 2025 at 2:21 AM Sam James <sam@gentoo.org> wrote:
>>> If you can test azanella's patches in an environment where you can
>>> reliably reproduce the bug, that would be very helpful indeed.
>> Will do if/when I figure out how to wrangle our build system to let
>> me use
>> local edits to glibc. But I've also attached a test that reproduces the
>> problem once every few thousand runs, at least when run on Google's
>> distributed test running system.
>> (It uses Google-isms, but all open source stuff, and all should be
>> easily
>> translatable if you want to avoid the dependencies.)
> If you could do the work to turn this into a plain ISO C + POSIX test case
> then that would be really useful for our efforts to add a regression test.