On Thu, 2007-01-04 at 17:16 -0500, seth vidal wrote: > On Thu, 2007-01-04 at 23:09 +0100, Terje Rosten wrote: > > * seth vidal > > | > > > > | I think I got it. > > | Terje, Tim, please take a look at the patch here: > > | > > https://lists.dulug.duke.edu/pipermail/yum-cvs-commits/2007-January/001255.html > > | > > | I did timing tests and it reduces the returnObsoletes call on my system > > | from 30s to 3s > > > > Yes, it's better, however 3.0.2 plus patch is still much slower than > > 3.0.1 on 'yum install xpdf' with warm cache. > > B/c 3.0.2 is performing a whole other type of check that 3.0.1 isn't. > Instead of just looking at all the obsoletes whole sale. It taking the > list of obsoletes, removing from them any package which is not of the > newest set, then performing the updates setup check. > > I'll see what I can do to trim it out a little further. > > > > The this command: > > > > $ time python yummain.py -y install xpdf > > > > I get: > > > > 3.0.1: 9 sec > > 3.0.2 + plus patch: 15 sec > > > > >From the profiling data I see lots of call to returnNewestByName, it's > > called over 8500 times. > right - that'd be once per package in the obsdict. I think I know how to > streamline that. > > > I'll let you know what I come up with.
okay - I just checked in another attempt. https://lists.dulug.duke.edu/pipermail/yum-cvs-commits/2007-January/001257.html can you tell me what you get from that with your profiler. The slow down shouldn't be from so many calls to returnNewestByName() -sv _______________________________________________ Yum-devel mailing list [email protected] https://lists.dulug.duke.edu/mailman/listinfo/yum-devel
