Re: svn commit: r18260 - trunk/subversion/libsvn_ra_serf

On 1/27/06, Greg Stein <gstein@lyra.org> wrote:
> On Fri, Jan 27, 2006 at 09:42:31AM +0100, Peter N. Lundblad wrote:
> >...
> > > - /* todo: create a subpool? */
> > > - serf_sess->pool = pool;
> > > + apr_pool_create(&serf_sess->pool, pool);
> >
> > Should be svn_pool_create, but why do you need a subpool?
>
> Right. Unless it is going to be cleared at some points, then a subpool
> isn't going to be all that helpful.

The key for the subpool is to have serf's socket be in a separate pool
from the main session pool. Without that (i.e. everything in one
pool), the underlying socket gets cleared before the serf cleanup
routines are called and it leads to a segfault. Placing serf in its
own subpool resolves this problem.

> If it gets misspelled, then it simply won't work. The failure mode is
> direct and obvious, so a symbolic conatant isn't going to improve
> things.

I'm tossing ideas around about how to better handle properties. Not
sure exactly how it'll play out though. Stay tuned on that front. --
justin

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Fri Jan 27 18:28:46 2006