Re: Another 1.4 release critical bug

On 8/3/06, Garrett Rooney <rooneg@electricjellyfish.net> wrote:

> Hmm. I'm not remembering that sort of thing, but perhaps it was
> before I got intimately involved with the codebase...
>
> Anyone else remember the details of this problem?

Ok, yeah, I dug around some more and yes, I now agree, it's not thread safe.

The pools code has a bunch of locking, but it's only ever used when
the pools fall back to getting memory out of an apr_allocator_t, the
default behavior, when they don't have to do that, is totally not
thread safe. There's even a comment in apr_pool_check_integrity about
how the integrity checking (specifically the thread ownership
checking) falls down if someone searched out the top level parent pool
and used that ;-)

New this was too easy to be correct... Ok, back to the drawing board...

-garrett

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Thu Aug 3 16:04:20 2006