D6255: copies: calculate mergecopies() based on pathcopies()
martinvonz (Martin von Zweigbergk)
phabricator at mercurial-scm.org
Mon Apr 29 18:56:19 EDT 2019
martinvonz added a comment.
In https://phab.mercurial-scm.org/D6255#91965, @martinvonz wrote:
> In https://phab.mercurial-scm.org/D6255#91851, @martinvonz wrote:
> > I'll spend a bit more time to see if I can figure out why pathcopies() and mergecopies() walk file ancestor differently. The way mergecopies() does it is faster, so I'l see if I can use that for pathcopies() too. https://phab.mercurial-scm.org/D6274:: can still be queued if anyone has time.
> I thought I was done with that after finding some bugs in mergecopies(). I thought fixing those would make mergecopies() as slow as pathcopies(), but that still doesn't seem to explain it :( Maybe I'll spend even more time on this tomorrow.
The biggest difference turned out to come from the `isintruducedafter()` that I mentioned earlier. I'd be fine with removing that, but we can discuss that after this patch is landed. I think it's an improvement to make pathcopies() and mergecopies() more consistent anyway.
While investigating differences between pathcopies() and mergecopies(), I noticed some other differences and I've added tests for them. As you can see in this patch, some of them are now fixed.
To: martinvonz, #hg-reviewers
Cc: marmoute, mjpieters, mercurial-devel
More information about the Mercurial-devel