Re: [Condor-devel] Crashing gdb in Create_Process

[Brian Bockelman <bbockelm@xxxxxxxxxxx>]

On Jun 3, 2011, at 11:23 AM, Alan De Smet wrote:

> Brian Bockelman <bbockelm@xxxxxxxxxxx> wrote:
>> This is laid out in #2208.  It seems that gdb will assert on Condor compiles that use pthreads and the clone call, as Condor confuses pthreads by creating a new thread.  This happens whether you ask gdb to follow the parent or child.
> 
> After reading the ticket, I'm not clear: to what extent is this
> gdb's fault?  Is this a matter of gdb having a bug, and we should
> generate a simple test case and submit it upstream for a fix?  Or
> is it a matter of gdb simply cannot do anything better and we
> need to place nicely?
> 

It's a fairly subtle interaction between gdb and glibc.  It's not clear who's at fault.

However, anything linked with pthreads that does clone+exec should be able to trigger the assert in GDB.

> Either way, it would be nice to be able to debug the common code
> path in Condor.  Thanks for digging into it!
> 


Yeah - it was starting to get on my nerves.

Brian