You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by Greg Stein <gs...@lyra.org> on 2001/03/30 14:10:21 UTC

grr. DB crashing.

So... Berkeley DB is crashing on me. (wah!)

When the FS gets cleaned up at pool-destruction time, it goes to close the
"nodes" table. Inside DB, it destroys a list of cursors. Something down in
libc barfs when one of those cursors is free'd.

Sigh.

libc's free() function has a pthread mutex in it, which it tries to lock. It
is that lock which bombs. Annoying.

I've built a debugging version of DB, but that isn't helping right now. It
doesn't say much more since the issue is in libc. It doesn't appear to be a
multiple free, but it doesn't look like it would adequately test that, so
I'll add some testing code into DB for that.

Hmm. Gonna have to do some more work, I guess. Anybody else seeing this?

This has me stuck...

Off to bed now.

Cheers,
-g

-- 
Greg Stein, http://www.lyra.org/

Re: grr. DB crashing.

Posted by Karl Fogel <kf...@collab.net>.
Karl Fogel <kf...@collab.net> writes:
> Just an update, Ben and I traced this a little farther today, but
> unfortunately we both have to run before we could finish.  However, I
> think it's safe to say that the bug will meet an early end, sometime
> this weekend. :-)

Chip, don't let us put you off from debugging!  We'll be somewhat out
of commission this week; if you can find time to debug this, that
would be awesome.

Re: grr. DB crashing.

Posted by Karl Fogel <kf...@collab.net>.
Greg Stein <gs...@lyra.org> writes:
> So... Berkeley DB is crashing on me. (wah!)
> 
> When the FS gets cleaned up at pool-destruction time, it goes to close the
> "nodes" table. Inside DB, it destroys a list of cursors. Something down in
> libc barfs when one of those cursors is free'd.
> 
> Sigh.
> 
> libc's free() function has a pthread mutex in it, which it tries to lock. It
> is that lock which bombs. Annoying.
> 
> I've built a debugging version of DB, but that isn't helping right now. It
> doesn't say much more since the issue is in libc. It doesn't appear to be a
> multiple free, but it doesn't look like it would adequately test that, so
> I'll add some testing code into DB for that.

Just an update, Ben and I traced this a little farther today, but
unfortunately we both have to run before we could finish.  However, I
think it's safe to say that the bug will meet an early end, sometime
this weekend. :-)

-K

Re: grr. DB crashing.

Posted by Karl Fogel <kf...@collab.net>.
Greg Stein <gs...@lyra.org> writes:
> So... Berkeley DB is crashing on me. (wah!)
> 
> When the FS gets cleaned up at pool-destruction time, it goes to close the
> "nodes" table. Inside DB, it destroys a list of cursors. Something down in
> libc barfs when one of those cursors is free'd.
> 
> Sigh.
> 
> libc's free() function has a pthread mutex in it, which it tries to lock. It
> is that lock which bombs. Annoying.
> 
> I've built a debugging version of DB, but that isn't helping right now. It
> doesn't say much more since the issue is in libc. It doesn't appear to be a
> multiple free, but it doesn't look like it would adequately test that, so
> I'll add some testing code into DB for that.
> 
> Hmm. Gonna have to do some more work, I guess. Anybody else seeing this?

Yes, I am seeing this the last couple of days, and also have
reproduction recipes, as it sounds like you do.  I hadn't started
debugging it yet (too many burners, not enough cooks :-) ); it sounds
like you've already traced it a bit further than I had.

> This has me stuck...
> 
> Off to bed now.

Any more news since you went to bed? :-)

(Ben and I were planning to debug this today, so we've love to know
any leads you have...)

-K