On Wed, 2004-04-28 at 03:56, Branko Èibej wrote:
> A consequence of removing the baton and checks from compose_window is
> that now compose_handler only checks B->src_ops when it sets up the
> initial baton, but those are never checked later. Which means that if a
> windoe in the middle of the delta uses no source ops, we'll still do the
> combination. Given your cnahges in this commit, I think
> svn_txdelta__compose_windows should check B->src_ops. I'd also change
> the allocation guarantee slightly, so that compose_windows can return B
> without having to copy it to the pool (the compose_handler
> implementation already assumes so, but that't not documented).
Please see http://www.contactor.se/~dast/svn/archive-2004-04/1183.shtml
if you missed it the first time around.
I do think that you're correct, and I misread the FS code I was
modifying. But I think the right fix is more along the lines of:
Index: subversion/libsvn_fs/reps-strings.c
===================================================================
--- subversion/libsvn_fs/reps-strings.c (revision 9551)
+++ subversion/libsvn_fs/reps-strings.c (working copy)
@@ -189,6 +189,7 @@
svn_pool_destroy (cb->window_pool);
cb->window = composite;
cb->window_pool = composite_pool;
+ cb->done = (composite->sview_len == 0 || composite->src_ops == 0);
}
else if (window)
{
Does that seem reasonable?
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Wed Apr 28 21:04:53 2004