Re: svn commit: r25869 - in trunk/subversion: include/private libsvn_fs_fs tests/libsvn_fs

On Wed, Sep 19, 2007 at 09:26:02AM -0700, David Glasser wrote:
> On 9/19/07, Blair Zajac <blair@orcaware.com> wrote:
> > Didn't think of locking transaction-current. So we could definitely
> > do that.
>
> Yes, if it is feasible to make the transaction-current and commit
> locks separate (and I don't see why it wouldn't be...) I would
> definitely be +1.
>

We can (and probably should: if we don't fix it for 1.5, we won't be
able to fix it easily thereafter).

Unfortunately, I forgot that we also need to create and manage a mutex
for cross-thread synchronisation, so we'd need to duplicate a fair bit
of the current write-lock code. Anyone see an easy way to abstract that
away so that we don't need to copy and paste everything? (bonus
question: anyone see a way to abstract it into a file that's _not_
fs_fs.c, which is ~180KB and growing?)

Regards,
Malcolm