Branko Čibej <brane@xbc.nu> writes:
> >RA->open() calls that internal function over and over, until it
> >doesn't return error anymore, i.e. it actually succeeded in finding a
> >BDB environment to open. That's exactly one opening of the BDB
> >environment... one per RA->open() call.
> >
> >
> Well, all I can say is that on HP-UX, diff, merge, switch, etc. break
> when we try to mmap an already-mmapped BDB environment file while
> opening the repo a second time in svn_ra_local__split_URL. Where does
> ra_session2 come from, if not from there?
Yes, for diff, merge, switch, etc, we deliberately open exactly two ra
sessions to the same URL, as I explained previously. In ra_local
terms, this means opening the same BDB environment twice, and I
understand that this breaks HP-UX, and I understand that you don't
like it. :-)
But in the issue, Brandon points out that the BDB files seem to be
open 80 times each during a large merge, which seems mighty fishy to
me. All I'm trying to point out here is that in *theory*, we only
open the environment exactly twice, not 80 times.
Of course, maybe this is some weird BDB implementation thing. If I
open an svn filesystem, and then call svn_fs_file_contents()
repeatedly on different paths, is that going to open the DB files over
and over? I really don't know.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Tue Mar 4 22:11:42 2003