On Jan 12, 2005, at 4:44 PM, Ben Collins-Sussman wrote:
> Hmm, it's issue 1167:
>
> http://subversion.tigris.org/issues/show_bug.cgi?id=1167
>
> Apparently it's been floating out in "unscheduled" space forever.
>
> I originally thought someone had tried and failed to do it,
> discovering it was very hard. But talking to my officemates, that
> seems not to be the case. I think it's just that nobody's gotten
> around to it. kfogel says that some developers feel that the issue
> should probably be part of a "larger rewrite" of the svn:externals
> feature, so maybe that's why it's not had much inertia.
>
> In any case, cmpilato thinks it would be pretty easy for someone to
> implement. It's just a matter of teaching svn_client_commit() to loop
> over the list of externals and make more commits.
Is it correct to assume that "update" does exactly the same sort of
loop, so the code to do the loop in general (probably a small part of
the entire task) is already present somewhere?
> (Since the externals are often from different repositories, we
> couldn't possibly make one giant atomic commit.)
Would it be difficult o make sure that everything from the same
repository was part of a single atomic commit, so it would be the
theoretical minimum of 1 commit per repository? I think that should be
the goal, even if it doesn't happen in the first implementation.
Scott
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Thu Jan 13 05:06:35 2005