Greg Hudson <ghudson@MIT.EDU> writes:
> On Mon, 2005-04-25 at 13:12, Philip Martin wrote:
>> > If I have a locally-modified file "foo" and I receive an update which
>> > adds "newfoo" copied from foo, I propagate the local mods, and now have
>> > them in two places. But if I check in my local mods before the update
>> > (no conflict), then the modifications are never propagated to newfoo.
>
>> Well that's another thing I think should change. If I merge a change
>> that copies foo.c to bar.c I don't really want the merge to copy bar.c
>> from the merge source, I want it to copy it from my working copy
>> source.
>
> You seem to have switched tracks from an "svn update" scenario to an
> "svn merge" scenario, after previously saying that you were primarily
> concerned with "svn update" scenarios before. That's very confusing.
Yes, I did. Sorry. I'm probably a bit confused as well.
> If I try to apply your new proposal to "svn update" scenarios like the
> one quoted with "> >" above, I get bizarre consequences. For instance,
> when I added editorp.c to subversion/libsvn_ra_svn (copied from
> editor.c), it sounds like anyone doing an "svn update" would wind up
> with whatever version of editor.c they previously had checked out copied
> to editorp.c, which would represent a completely fabricated and
> unintended local mod. (I think we can take as axiomatic that a sequence
> of only "svn co" and "svn up" operations should never produce a working
> copy with local modifications.)
I don't think conversion of the copyfrom source URL should apply to
update.
I still think it's possible that update should cause local mods to be
copied. It often does the right thing during a move, and it's easy to
revert if the local mods are inappropriate.
To take your editorp.c example, the difference between local mods
being copied and non-local mods not being copied can be explained as
the consequence of a race. (Between one user making the copy and
another user committing mods to the copy source.) If you had noticed
a commit to editor.c while you were working on the commit that
produced editorp.c then you would probably have considered transfering
those changes to editorp.c yourself. If you didn't notice such a
commit then we would be likely to just put it down to a race.
--
Philip Martin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon Apr 25 19:50:02 2005