Re: cvs branch tag performance - fix offering

[Mark D. Baushke]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Adrian West <address@hidden> writes:

> I  have a fairly large project being managed with cvs very
> successfully (thanks).
> A problem has been the performance of branch tag creation, which was
> taking half an hour or more.
> 
> I have a fix for this now. It's not expecially elegant, and will only
> affect large projects using lots of symbols, so submitting it directly
> may not be appropriate. It if for 11.14, which is what we happen to be
> using at the moment.
> 
> If it will help you the details, and some analysis of our last few
> months operation with it are at:
> 
> http://www.adrianwest.org/cvs_perf_fix/cvs_perf_fix.htm

Such a patch will not be considered for cvs 1.11.x as we do not want to
destabalize that release, but some kind of performance improvement may
be possible for cvs 1.12.x.

I only took a quick look. This patch would likely have problems for
users that wish to use CVSup with CVS_LOCAL_BRANCH support (added to cvs
1.12.9). I suspect you may wish to have the starting magic branch number
for your search be something that is configurable in the CVSROOT/config
file.

Your patch also brings up an interesting concern... what happens when
you have enough branches that rev_num wraps negative? It has mostly been
considered that 2^30 branches for a given revision would be enough. If
that is not true, it would be good to know it...

        -- Mark
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (FreeBSD)

iD8DBQFBpOQK3x41pRYZE/gRAmQbAKCWUBmDalHSodLI3Y8akBXFwnRoAwCg2w0G
89OjF3thZ07zA9Ffcida+yE=
=hegB
-----END PGP SIGNATURE-----