Re: [RFC PATCH glibc 24/34] hurd: Only check for TLS initialization inside rtld or in static builds

[Sergey Bugaev]

On Wed, Apr 12, 2023 at 12:23 AM Samuel Thibault
<samuel.thibault@gnu.org> wrote:>
> Sergey Bugaev, le mar. 11 avril 2023 23:27:35 +0300, a ecrit:
> > On Tue, Apr 11, 2023 at 9:57 PM Samuel Thibault <samuel.thibault@gnu.org> 
> > wrote:
> > > Had you actually tested it on i386? It seems to be breaking the
> > > testsuite completely. I would expect that a submitted patch series has
> > > gone through the testsuite.
> >
> > Ouch!
> >
> > I have tested that it works on i386, as in I was able to run bash and
> > apt with it. I'll re-check whether it (still) works,
>
> Yes, some things work, but some others don't :)
>
> > I have not run the testsuite, because:
> > * I'm cross-compiling
>
> You can probably push your tree to a box where you can compile & run?
>
> > * I never managed to run it to completion on the Hurd in the first
> > place... but maybe you have already fixed this.
>
> I have marked various tests as unsupported, yet.

Alright, this went much as expected. First it took an eternity to
compile (SMP support can't come soon enough), and second, the testuite
has brought the system into a bad state and didn't complete. In
particular: it managed to terminate my SSH session several times (I
don't know how), and eventually it ran into that networking stack
deadlock (the only way out of which seems to be reboot-hurd on the
console). So I still can not run the testsuite on my end.

But the tests that did run seem to mostly have worked. There were some
failures here and there (I don't know which ones are expected), but
nothing as bad as every single program crashing because it starts
using TLS before it's set up due to already having switched to libc.so
functions.

Could you please point me to a specific test case (and not just "run
the whole thing") that started failing because of this change, and
teach me how to run just that one? I have read [0], but that says, "To
test just one test you have to have already run the entire testsuite",
which is a non-starter.

[0]: https://sourceware.org/glibc/wiki/Testing/Testsuite#Testing_just_one_test

Side note, I don't know why it's glibc's test suite that brings the
system to its knees. glib's, which is also quite comprehensive,
completes just fine. Would it have been easy for me to run the full
test suite, I would surely do that before submitting any patches. But
it's not.

Sergey