fix cond_race... was RE: src/winsup/cygwin ChangeLog thread.cc thread.h ...

Robert Collins robert.collins@itdomain.com.au
Wed Oct 24 21:12:00 GMT 2001


> -----Original Message-----
> From: Jason Tishler [ mailto:jason@tishler.net ]
> Sent: Thursday, October 25, 2001 10:47 AM
> To: Robert Collins
> Cc: cygwin-developers@sourceware.cygnus.com
> Subject: Re: fix cond_race... was RE: src/winsup/cygwin ChangeLog
> thread.cc thread.h ...
> 
> 
> Rob,
> 
> On Mon, Oct 22, 2001 at 12:52:14PM +1000, Robert Collins wrote:
> > right, then try sleep (1) where I have 0 :}
> 
> AFAICT, the test case still hangs in the same place or takes 
> a really long
> time to complete.  Even if I try to be tricky (like the aforementioned
> Python code) and alternate between Sleep(0) and Sleep(1).
> 
> Any other ideas?

Add logging code (such as system_printf to the TimedWait method
immediately before the WFSO call. Add similar code to the Signal method
immediately before the PulseEvent.

The log detail should include the thread id (I _think_ thread_printf
includes this and logs to strace) and a timestamp. 

Process the log file to show both events, in time order, and see if
naything pops out. If not forward me the log. 

Do you have sshd on your machine? Or can you bzip2 and mail me a
build+source tree that shows this symptom and is ready-to-go?

Rob



More information about the Cygwin-developers mailing list