mercurial/crew at 10264: 4 new changesets (3 stable): eb243551cbd8 is probably wrong

Matt Mackall mpm at selenic.com
Fri Jan 22 13:20:49 CST 2010


On Fri, 2010-01-22 at 19:35 +0100, Patrick Mézard wrote:
> > http://hg.intevation.org/mercurial/crew/rev/eb243551cbd8
> > changeset:   10262:eb243551cbd8
> > branch:      stable
> > user:        Matt Mackall <mpm at selenic.com>
> > date:        Tue Jan 19 22:20:05 2010 -0600
> > summary:     copies: speed up copy detection
> 
> Trying to rebase 15 or 20 revision involving at least one file which
> was moved twice in difference locations, I had a really strange merge
> request which really made no sense. So I backed out eb243551cbd8 and
> retried the rebase and this time I got something like (shell output is
> a little garbled):

Hmmm, can you try reducing it to a test case? There's one minor issue
I'm aware of: the code does definitively determine the files it marks as
copies are related at some point equal to or after the ancestor, but it
doesn't actually determine that they are both the same file in the
ancestor. Because of the way rebase fiddles with ancestor(), interesting
things might happen.

Also "reverse merges" have issues with copies that this new code might
make worse.

-- 
http://selenic.com : development and support for Mercurial and Linux




More information about the Mercurial-devel mailing list