Ben Collins-Sussman <sussman@collab.net> writes:
> Wow, we're definitely closing in on this one! I can reproduce the
> wedge bug with 0.28 HEAD, but not at all with 0.27. (I believe Karl
> just reproduced the wedge *exactly* as I did.)
Almost exactly, yep! The only differences are: my failures happen
very soon after the checkouts start, and the checkouts don't hang for
a long time before the errors happen. (My checkouts are also
exceedingly slow when they're still running, though).
Details:
I'm using Subversion 0.28.1 (built from the tarball) to do 3
simultaneous checkouts from http://localhost/. The repository is a
copy of revision 6917 of the Subversion repository at svn.collab.net.
Each checkout gets a little way and then dies. The first one to die
errors with "cannot allocate memory":
A /home/kfogel/src/wedge/checkouts/wc1/trunk/www/project_issues.html
A /home/kfogel/src/wedge/checkouts/wc1/trunk/www/images
subversion/libsvn_ra_dav/util.c:597: (apr_err=160029)
svn: Berkeley DB error
svn: PROPFIND request failed on '/repos/!svn/bc/6874/trunk/www/images'
subversion/libsvn_ra_dav/util.c:331: (apr_err=160029)
svn:
Berkeley DB error while fetching string length for filesystem /home/kfogel/src/wedge/repos/db:
Cannot allocate memory
The other two then fail right away with DB_RUNRECOVERY errors:
A /home/kfogel/src/wedge/checkouts/wc2/trunk/www/project_issues.html
A /home/kfogel/src/wedge/checkouts/wc2/trunk/www/images
subversion/libsvn_ra_dav/util.c:331: (apr_err=160029)
svn: Berkeley DB error
svn:
Berkeley DB error while beginning Berkeley DB transaction for filesystem /home/kfogel/src/wedge/repos/db:
DB_RUNRECOVERY: Fatal error, run database recovery
And
A /home/kfogel/src/wedge/checkouts/wc3/trunk/www/getting_subversion.html
A /home/kfogel/src/wedge/checkouts/wc3/trunk/www/subversion-diagram.png
subversion/libsvn_ra_dav/util.c:331: (apr_err=160029)
svn: Berkeley DB error
svn:
Berkeley DB error while beginning Berkeley DB transaction for filesystem /home/kfogel/src/wedge/repos/db:
DB_RUNRECOVERY: Fatal error, run database recovery
Now I'm going to go make sure this *doesn't* happen with 0.27 checking
out from a format-1 repository, so it's definitely narrowed down to
something about 0.28.
-Karl
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Fri Aug 29 18:31:27 2003