On Mon, 26 Nov 2007, David Glasser wrote:
> On Nov 26, 2007 2:31 PM, Daniel Rall <dlr@collab.net> wrote:
> > On Thu, 22 Nov 2007, David Glasser wrote:
> >
> > > On Nov 22, 2007 3:27 AM, <dlr@tigris.org> wrote:
> > > > Author: dlr
> > > > Date: Thu Nov 22 00:27:04 2007
> > > > New Revision: 27979
> > > >
> > > > Log:
> > > > When a new-enough version of SQLite is detected at 'configure'-time,
> > > > employ its thread-safety runtime verification.
> > >
> > > Should we only do this if APR_HAS_THREADS?
> >
> > Theoretically, doesn't that depends on whether r27979 is relevant in a
> > situation where a forking consumer of libsvn_fs* uses libsqlite in
> > conjunction with shared memory?
>
> I'm confused. Sqlite uses shared memory?
Dunno. But an external consumer might put a Subversion object which
referred to some sqlite object into shared memory and use it without
any sort of mutex (e.g. assuming that it's thread-safe). *shrug*
- application/pgp-signature attachment: stored
Received on Tue Nov 27 00:19:57 2007