[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

RE: [Subclipse-dev] Performance regression in Quick Diff (0.9.32)

From: Martin Letenay <mle_at_whitestein.com>
Date: 2005-07-29 09:40:30 CEST

> > Hi All,
> > I have just tried the new release and found a pretty massive
> > performace regression for remote repositories. It is hitting the
> > server to get the contents of the reference document (once only
> > though). I didn't spot it before because i use a local repo.
> >
> > It looks like we actually use a remote resource in our sync info to
> > represent the base revision (like CVS does). I have
> attached a patch
> > that fixes the performance problems for me.
> > It will now use a base resource, so getContents will be a local
> > filesystem read now. It also removes some redundant throws clauses.
> >
> > This is only going to be a problem for people who connect to a
> > repository over a slow connection (such as the internet).
> >
> > I would have checked it in, but I know very little about the sync
> > stuff and don't want to make things worse...
> >
> > cheers,
> > Brock
>
> Well, I can take a closer look at consequences of your patches.
>
> Yes, I've noticed some lag when comparing changed resources
> from synchronize view.
> I did not put much thought on it, just a very shallow observation was:
> - it was slow in 0.9.31 when I joined the project.
> - after I did first set of patches, I've observed an
> unintentional positive side-effect there, there was no lag
> anymore. But it was not my intention, it just happened.
> - when the sync code was rewritten for 0.9.32 the lag is there again.
>
> I'll investigate this issue tonight or tomorrow ...

Hello Brock,

Just a status update:
I examined your patches and their consequences and they are of course
correct.
What just happened is that your fixes inspired me to take a deeeeper look
into
surrounding classes and then I found myself refactoring most of the
IResourceVariant
classes and the way sync infos are constructed.
I expect I might get finished by this weekend.

The expected result is again some increase in performance and maybe also
some behavioural improvement.
It's a bigger stuff, so I might need your help with testing then ...

Regards,
Martin
Received on Fri Jul 29 17:40:30 2005

This is an archived mail posted to the Subclipse Dev mailing list.