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

Re: Performance regression with reverse merge

From: Mark Phippard <markphip_at_gmail.com>
Date: Sun, 1 Mar 2009 15:34:49 -0500

On Sun, Mar 1, 2009 at 2:52 PM, Stefan Fuhrmann
<stefanfuhrmann_at_alice-dsl.de> wrote:
> On Sunday 01 March 2009 18:33:48 Mark Phippard wrote:
>
>> > $ uname -a
>> > Linux opti 2.6.24-23-generic #1 SMP Mon Jan 26 00:13:11 UTC 2009 i686
>> > GNU/Linux $ svn --version
>> > svn, version 1.4.6 (r28521)
>>
>> Since you are calling this a 1.6 regression, why are you not comparing
>> against 1.5.6?
>
> Sorry! My bad; I wanted the post out quickly and had no
> 1.5 on my machine.
>
> 1.5.6 numbers:
>
> $ time ./svn merge -r 15456:15455
> http://tortoisesvn.tigris.org/svn/tortoisesvn/trunk ~/TSVN3
> --- Reverse-merging r15456 into
> '/home/stefan/TSVN3/src/TortoiseProc/RevisionGraph/RevisionGraphDlg.cpp':
> U    /home/stefan/TSVN3/src/TortoiseProc/RevisionGraph/RevisionGraphDlg.cpp
> --- Reverse-merging r15456 into
> '/home/stefan/TSVN3/src/TortoiseProc/RevisionGraph/RevisionGraphDlg.h':
> U    /home/stefan/TSVN3/src/TortoiseProc/RevisionGraph/RevisionGraphDlg.h
>
> real    6m52.263s
> user    0m2.220s
> sys     0m0.564s
>
> So, this is basically the same as for 1.6.
> Obviously, I never needed to undo revisions
> for quite a long time.

So then there is no release blocking 1.6 regression here. Agree?
Mainly we need to look into why this is slow and see if it can be
improved.

pburba made an optimization so that if your WC is at a single uniform
revision that it does some performance stuff that made a big
difference. I wonder if it is not implemented for reverse merge? Can
you try merging a single revision from a branch to see if the
performance is similar or better?

>> > $ svn co http://tortoisesvn.tigris.org/svn/tortoisesvn/trunk -r15533
>> > ~/TSVN > /dev/null
>>
>> You might have mentioned how long it takes this checkout to run?  I am
>> running checkout from a box in the CollabNet network and it is still
>> taking forever.
>
> Hm. Around 4 minutes here, including externals.
> I didn't want to clutter my post with numbers
> that I didn't consider relevant for merge performance.

That is strange. I was basically LAN-connected to tigris and I think
checkout took between 15 and 20 minutes. It was long enough that I
decided it was taking a long time and that I'd also run the tests on a
different system (Windows). Managed to connect to that box, run the
diff tests, find and install the timethis.exe software and get most of
the "src" checkout done.

-- 
Thanks
Mark Phippard
http://markphip.blogspot.com/
------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=462&dsMessageId=1252275
Received on 2009-03-01 21:35:09 CET

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

This site is subject to the Apache Privacy Policy and the Apache Public Forum Archive Policy.