Re: Locks continued...

[Tobias Brox]

[raptor - Wed at 12:36:34PM +0200]
> 
> 1. CVS walks and see test1 and test2 has to be commited 'cause they are
> changed
> 2. test1 and test2 are commited
> 3. CVS call log-info script
> 4. log-info is started and it forks process to do the main-working copy
> update
> 5. Meanwhile CVS continues to check blah-dir and see that blahtest has to be
> updated too
> 6. BUT we are updating at the moment and CVS now stops waiting for lock to
> be freed (which doesn't happen)

Hm ... I never got into that problem, because I checked in each directory
locally, and had some seconds break between each directory.

The lock should be freed, though.

Also, the locks /should/ (though I haven't studied this) work on an
per-directory-basis, thus updateing one directory while committing to a
subdirectory shouldn't be harmful.

I didn't study your code much, but it seems like you don't use any arguments
to it?  Use the %s argument in the loginfo file, parse it in the script, so
that it will do "cvs update mydir/myfile" instead of "cvs update"?

Check out my cvs-tripwire-script to see how I've done the parameter parsing
there - I'll attach it privately to you.  I did the parsing a bit robust, so
that filenames with spaces might pass.

-- 
Unemployed hacker
Will program for food!
http://ccs.custompublish.com/