Ben Collins-Sussman <sussman@collab.net> writes:
> Ben Collins-Sussman <sussman@collab.net> writes:
>
> > Ben Collins-Sussman <sussman@collab.net> writes:
> >
> > > 0.28 experiment: using svn's own repository (HEAD=6917, format=2)
> > >
> > > * 3 simultaneous checkouts caused repos wedge after 40 seconds or so:
> >
> > Okay, so I removed r6837 from my 0.28 HEAD tree and rebuilt.
> >
> > When I repeat the experiment, here's the unintelligible mess that
> > happens:
>
> Hm, nevermind. I just re-ran the test again, and all three checkouts
> died almost simultaneously, about 20 seconds in: two with
> 'runrecovery' error, the third with 'invalid xml' parse error. This
> is what Karl and I were seeing with plain old 0.28 anyway.
>
> So it looks like r6837 is pretty meaningless, and not a culprit.
If you get really desperate, try reverting revisions 6774 and 6809.
These made svn_fs_dir_entries() more costly (by shifting cost away
from callers of that function) -- but I wrote the code in a
lock-sensitive fashion, so I don't expect it to turn up much.
Also, it is important to remember that the svn.collab.net repository
was locking up (albeit not as frequently) with 0.27.0, too -- I have a
copy of a borked repos that dates back to August 14.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Fri Aug 29 19:53:19 2003