[Gluster-devel] netbsd regression update : cdc.t

[Krishnan Parthasarathi]

----- Original Message -----
> Krishnan Parthasarathi <kparthas at redhat.com> wrote:
> 
> > On gdb'ing into one of the brick process, I see the following backtrace.
> > This is seen with other threads in the process too. This makes it difficult
> > to analyse what could have gone wrong. Is there something I am missing?
> 
> Obviously frame 4 and higher are irrelevant, but you have glusterfs code
> in frame 3, with arguments, file and line. I guess if you type frame 3
> you get the code.
> 
> This posix_health_check_thread_proc() is probably spawed by
> pthread_create() and therefore has no caller.

Hmm. I didn't put my point clearly. The backtrace I chose to demonstrate the
problem is not representative. My bad.

The following threads have only the top-most frame. Is that expected?

<snip>
[Switching to thread 8 (LWP 8)]
#0  0xbb3c84e7 in ___lwp_park60 () from /usr/lib/libc.so.12
(gdb) bt
#0  0xbb3c84e7 in ___lwp_park60 () from /usr/lib/libc.so.12
(gdb) t 9
[Switching to thread 9 (LWP 7)]
#0  0xbb3c84e7 in ___lwp_park60 () from /usr/lib/libc.so.12
(gdb) bt
#0  0xbb3c84e7 in ___lwp_park60 () from /usr/lib/libc.so.12
(gdb) t 10 
[Switching to thread 10 (LWP 6)]
#0  0xbb3c84e7 in ___lwp_park60 () from /usr/lib/libc.so.12
(gdb) bt
#0  0xbb3c84e7 in ___lwp_park60 () from /usr/lib/libc.so.12
(gdb) t 11
[Switching to thread 11 (LWP 5)]
#0  0xbb3c84e7 in ___lwp_park60 () from /usr/lib/libc.so.12
(gdb) bt
#0  0xbb3c84e7 in ___lwp_park60 () from /usr/lib/libc.so.12
(gdb) t 12
[Switching to thread 12 (LWP 4)]
#0  0xbb3c84e7 in ___lwp_park60 () from /usr/lib/libc.so.12
(gdb) bt
#0  0xbb3c84e7 in ___lwp_park60 () from /usr/lib/libc.so.12

</snip>

On a related note, how do I identify the thread processing network
events via event_dispatch_poll? I wanted to make sure that none of the
above threads could possibly have poll(2) in its backtrace.
If we can assert that none of the threads are waiting on poll(2), then
it explains what we observe.