Re: pulseaudio mutex bug reported
On 24/11/11 04:06, Bob Tracy wrote:
> On Tue, Nov 22, 2011 at 04:12:53PM -0600, Bob Tracy wrote:
>> Search for #649641: pulseaudio mutex issue on Alpha architecture.
> The workaround mentioned in the report for ARM also works for Alpha.
> I'll leave it to a package maintainer to explain why :-).
I bet that it is a kernel issue. What kernel were you running when you
saw the mutex issues?
I have been running Debian kernel 3.1.1-1 (well, with a small tweak so
that I could get it built) with an up to date unstable installation on
the ES45 for a week. It is proving quite unreliable. While building
git with -j4 on the make, the system fell over in the git test suite.
First one CPU froze and became unresponsive, then later the oom-killer
was invoked and it all fell over. The freezes of a CPU are still
occuring, quite often with file I/O or trying to dismount a
partition/disc. I have just downgraded the kernel to 2.6.32 from the
Lenny backports distro (plus patches to get the missing syscalls, and
other fixes --- the same as what we are running on the buildds) and have
not seen a problem. Git test suite passed and have built a couple more
packages without problem.
I had also noted a couple of more months ago that building eglibc was
successful under kernel 2.6.38 but fails the mutex tests in the test
suite under kernel 2.6.39 or newer.
I intend to report this to the linux-kernel email list in the near
future once I have a bit more evidence.