On Wed, Apr 15, 2009 at 04:04:24PM +0530, Senthil Kumaran S wrote:
> (02:35:23 PM) stylesen: in
> "/subversion-dev/trunk/subversion/libsvn_wc/update_editor.c", line 5358I am
> trying to change the if check, to check for uuid and not ancestral relation
> (02:35:37 PM) stylesen: which fails on different url schemes
> (02:36:21 PM) gstein: you do not have enough information at that point in the
> code to fix that check
> The attached patch removes the check for scheme (of url) during a merge, which
> IMHO is not required, since we already know we are operating with the same
> repository
That statement seems to contradict what gstein said in the chat.
Which is correct?
> Index: subversion/libsvn_wc/update_editor.c
> ===================================================================
> --- subversion/libsvn_wc/update_editor.c (revision 37269)
> +++ subversion/libsvn_wc/update_editor.c (working copy)
> @@ -5355,12 +5355,9 @@
>
> new_URL = svn_path_url_add_component2(ent->url, base_name, pool);
>
> - if (copyfrom_url && ent->repos &&
> - ! svn_path_is_ancestor(ent->repos, copyfrom_url))
> - return svn_error_createf(SVN_ERR_UNSUPPORTED_FEATURE, NULL,
> - _("Copyfrom-url '%s' has different repository"
> - " root than '%s'"),
> - copyfrom_url, ent->repos);
> + /* ### TODO: Here we need to check copyfrom_url and ent->repos are from
> + ### the same repository by checking their UUIDs, which should be
> + ### trivial in wc-ng and avoids opening an ra_session here. */
Why is opening an RA session a problem?
Sure, performance will suffer, but if we need that information to make
the check correct, we have to get it to make the check correct.
Everyone already knows that wc-1 is slow, so the additional round trip
won't cause much surprises. It's the broken wc-1 design that is the
problem, and we can only try to work around it. That's all we can do.
> (02:39:32 PM) ehu`: stylesen: I don't think we should be fixing too many issues
> in wc-1, given that we're switching to wc-ng in the "near" future.
As Erik said, we have to stop chasing bugs in wc-1 at some point.
There are too many of them.
It does not make sense to waste too much time on it.
If there is some hard technical problem that prevents us from contacting
the repository during the check, I'd say we should make sure the check
is updated to use information available in wc-ng as soon as possible,
and forget about fixing the problem for wc-1. Otherwise, we go for the
RA session for now, and make the check use information available in
wc-ng as soon as possible, too.
Stefan
Received on 2009-04-15 14:31:30 CEST