[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Bug#673711: pthread_cond_timedwait returns immediately with ETIMEDOUT

reassign 673711 libc0.1
retitle 673711 pthread_cond_timedwait returns immediately with ETIMEDOUT
found 673711 eglibc/2.13-32
tags 673711 - upstream
user debian-bsd@lists.debian.org
usertags 673711 kfreebsd


On 20/05/12 23:56, Christoph Egger wrote:
> Nicolas Bourdaud <nicolas.bourdaud@gmail.com> writes:
>> http://freebsd.1045724.n5.nabble.com/pthread-cond-timedwait-broken-in-9-stable-from-JAN-10-td5487565.html

I know the symptoms are similar, but that sounded like a recent kernel
bug/change making the timers unreliable.  Whereas this seems more like a
familiar problem in GNU/kFreeBSD's glibc.

> I'm a bit skeptical hoqwever that this build failure is caused by a
> kernel bug in kfreebsd-9 ad the buildds (where this build failure
> happens) are running an 8.1 kernel which (according to the linked
> thread) is not affected.

Indeed, I tried Nicolas' testcase (thanks for that) which showed issues
on kfreebsd-i386 8.3-1-686 8.3-2, kfreebsd 9.0-1-amd64 9.0-3;  and of
course this was an issue for the kfreebsd-* buildds which run 8.1 kernels.

> However there was (IIRC) indeed some change in pthread/libc code
> recently that might cause this issue.

Robert's librt/pthread patches went into eglibc 2.13-32 to try fix this
kind of problem, but a few packages are still showing problems like
this.  Perl, ruby1.8, ruby1.9.1, and the Python-based waf builder a few
packages use (like guitarix), all have issues kinda similar to this.

I find it interesting that this bug led to those glibc errors (double
free or corruption) in #673681 instead of a hang or simple test failure.

Steven Chamberlain

Reply to: