This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [pushed] Improve gdb.threads/attach-many-short-lived-threads.exp timeout handling
- From: Pedro Alves <palves at redhat dot com>
- To: Doug Evans <dje at google dot com>
- Cc: gdb-patches <gdb-patches at sourceware dot org>
- Date: Mon, 09 Feb 2015 23:06:22 +0000
- Subject: Re: [pushed] Improve gdb.threads/attach-many-short-lived-threads.exp timeout handling
- Authentication-results: sourceware.org; auth=none
- References: <1423225537-26694-1-git-send-email-palves at redhat dot com> <CADPb22Q8ScnOfeCPbnzEJQaY58Pc7px_6irrZmV3QhMTeOfwSQ at mail dot gmail dot com>
On 02/09/2015 09:55 PM, Doug Evans wrote:
> On Fri, Feb 6, 2015 at 4:25 AM, Pedro Alves <palves@redhat.com> wrote:
>> The buildbot shows that this test is still racy, and occasionally
>> fails with time outs on some machines. I'd like to get major issues
>> with load out of the way.
>
> Thanks for fixing this.
>
> btw, how often do you run the testsuite in parallel?
> I see such load related issues all the time that way.
Yeah, I run the testsuite in parallel all the time, with
-j8, on an i7-2620M (2 cores / 4 threads).
The attach-many-short-lived-threads.exp never fails
for me though. I've now got access to Sergio's build slave,
the one that generates the buildbot test results that show the
racy failures, and when I run that test manually, in a loop, I
never see it fail. OTOH, through build bot the test fails
very often... This was last thing last Friday, and I
haven't managed to get back to it yet. The test seems to
expose more bugs, but the build bot also shows it FAILing in
a very odd way sometimes, like failing to attach the very first
time.
> [e.g., checkpoint.exp, gdb-sigterm.exp, et.al.]
Yeah, checkpoint.exp forks a ton of processes, and sometimes I'll
hit the "ulimit -u (max user processes)" limit when running in parallel.
I bump that "ulimit -u 10000" in the shell that I run tests on,
and then it never fails.
The gdb-sigterm.exp one should be fixed, I hope, since:
https://sourceware.org/ml/gdb-patches/2015-02/msg00151.html
Thanks,
Pedro Alves