Subversion Newbie <subversionnewbie@yahoo.com> writes:
> Again I'm at it trying to get decent (?) response times in "svn
> update" on a large repository. I'm using svn 1.1.1 and FSFS
> (not BerkDB), on Linux, on a fairly fast machine with fast SCSI
> disks.
>
> I create a test repository made from approx 94000 files in 7200
> directories, total of about 1.6 GB. (Takes about 13 minutes to
> "svn import" the whole thing.)
>
> Next, I check out the entire contents into two areas ("work" and
> "www"). I modify a single file (index.html in one of the
> directories) from the "work" area and commit it with
> "svn commit -m 'changed index' -N index.html"
>
> Then, I go to the "www" area to update. I run
> "svn update -N index.html" and it takes 30 seconds to update
> that one file.
> My question is, shouldn't this be _much_ faster, given that I'm
> telling it exactly what file(s) to update? (That's my
> understanding of what the "-N" option does.)
>
> By the way, an "svn update" of the whole thing, without the
> "-N" option, takes about 1:44 minutes.
>
> Any ideas? Or am I doomed to a slow "svn update" time with a
> repository this large?
You're right, it should be faster, I think. Not sure what's causing
the slowness.
Could you file an issue for this, pointing to the URL to your original
message in the archives, and mentioning that this is *not* a duplicate
of issue #1831?
Btw, "-N" does not mean "I'm telling you exactly what files". It
means "Do not recurse into subdirectories". By passing a filename on
the command line, you already told it exactly what files. No -N
should be necessary. But, I'd be curious to know if you get the same
behavior without the -N. Also, how many other files and subdirs are
in that particular directory?
-Karl
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Tue Nov 23 01:38:45 2004