This is great debugging work.

If I'm reading the output correctly, you may be able to suppress this
error by setting { resolution_template }->{ enabled } to 0. in
glue.conf. Can you set this off and try again ?

From a first glance, the CPU overhead seems like an issue with a general
regular expression used to search for resolution statuses.

On Fri, 2008-01-18 at 15:41 -0600, Brown, Mike wrote:
> Kristis,
> 
> Setting Apache's Timeout to 6000 didn't have any effect.  I also tried
> setting KeepAliveTimeout to 150 from 15.  No effect either.
> 
> FWIW, the commit without scmbug enabled using the big log message is less
> than 10 seconds.
> 
> I added the "Devel::Trace" package to my Perl and modified the pre-commit
> file to log the info /tmp/log.pre-commit.  I've attached the files in case
> the execution point of Perl when it has hung helps you.  It looks like Perl
> hangs at line 476 in SCM.pm.
> 
> Regards,
> Mike
> 
> -----Original Message-----
> From: Kristis Makris [mailto:[EMAIL PROTECTED]
> Sent: Friday, January 18, 2008 12:13 PM
> To: Brown, Mike
> Cc: '[email protected]'
> Subject: RE: [scmbug-users] very large commit comments break scmbug +
> bugz illa
> 
> 
> On Fri, 2008-01-18 at 10:32 -0600, Brown, Mike wrote:
> > Commits work using a local repository on the server that runs our
> subversion
> > system.
> 
> Clearly Apache has something to do with this. 
> 
> I can see how it would be possible for Apache to kill the Subversion
> commit process because Scmbug is taking a while to gather the
> information it needs ? E.g. playing with some httpd.conf options like
> "Timeout" may help (perhaps setting Timeout to a really high value).
> 
> I still don't understand how/why the log messages don't make much sense.
> Perhaps Apache is trying to kill the children processes of the incoming
> connection, meaning killing 'svnlook' (issued by Scmbug) while the
> Scmbug process isn't dead yet and is waiting for input from the file
> stream it opened, waiting indefinitely since 'svnlook' is dead.
> Eventually Scmbug should be killed by Apache too, but perhaps it isn't
> because it is the connection thread -- or mod_svn_dav has not properly
> detected the Scmbug process was killed already (or something to that
> effect)
> 
> This is all speculation.
> 
> The default Timeout value in apache may be set to 300 (5 minutes). I can
> see how a commit process could take that much time. In my tests
> committing a  large message with libxml it took 3minutes -- so its
> close. Try setting Timeout=6000 and see if the commit succeeds within
> 1hr:20 minutes.
> 
> 

Attachment: signature.asc
Description: This is a digitally signed message part

_______________________________________________
scmbug-users mailing list
[email protected]
http://lists.mkgnu.net/cgi-bin/mailman/listinfo/scmbug-users

Reply via email to