Greg Ames wrote:
> 
> fyi.  I count 221 seg faults here.  Time to go back to a more reliable level of
> code on daedalus, once the workload lightens up.

Jeff and I found what was causing 2 out of the 3 dumps we've captured.  The
commit to http_protocol.c has the details.  I'm creating a build with:

* 2.0.35,
* the patch to save the socket input buffers for debugging,
* Cliff's patch for mmap cleanups, and
* the new http_protocol.c patch.

This will be called apache2.0.35d.  I expect to put it into production later
this evening, assuming it passes the usual tests.

My hunch is that dump 2 (the mmap cleanups) is less frequent.  

Greg

Reply via email to