Re: [PATCH] Fixing the FSFS corruption bug, again

On Tue, Oct 03, 2006 at 09:06:05AM -0400, Garrett Rooney wrote:
> Yeah, I just happened to miss that bit. Now that I've read it, I
> understand totally.
>

Excellent!

I committed a version of this patch (sans enormous lines) in r21738.

> >My only concern with the approach I've taken is the possibility of
> >accreting a collection of transactions that have neither been committed
> >nor aborted; these would lie around in memory in the same way they'd lie
> >around on disk. On the other hand, that shouldn't be that common a
> >problem, should it?
>
> I'm not sure there's a real work around for it anyway.
>

I agree; I can't see a simple way to solve this short of
reference-counting the shared transaction objects, and adding pool
cleanups to the svn_txn_t's pools. And even then, I'd be worried about
pool destruction ordering.

Regards,
Malcolm

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Tue Oct 3 18:04:57 2006