pulkit added a comment.
In https://phab.mercurial-scm.org/D4887#73793, @martinvonz wrote: > > I'm still fine with queuing this for now if you need this patch somewhat urgently even though I think the code in this patch and in https://phab.mercurial-scm.org/D4886 should eventually go away. Well, we do want to optimize this manifest thing before the upcoming release because it adds 3-4 minutes to some narrow-copy extending. I also agree that all these hacks should go away. I was thinking to implement something like widen_changegroup() which will generate the changegroup which we want instead of hacking into the original changegroup code. INLINE COMMENTS > changegroup.py:1051 > prunednodes = self._prunemanifests(store, nodes, commonrevs) > if tree and not prunednodes: > continue > I don't think this is quite the right approach. If we're using treemanifest > and widening from {foo/, bar/} to {foo/, bar/, baz/}, then we'd still be > sending manifests for all those directories (and their subdirectories, > right)? We should only have to send manifests for baz/ (and subdirectories). @martinvonz IIUC because of this and condition in line 1041 above, we won't send the manifest for all those directories and we will be sending manifest for baz/. The filematcher here is the differencematcher so foo/ and bar/ does not match it. REPOSITORY rHG Mercurial REVISION DETAIL https://phab.mercurial-scm.org/D4887 To: pulkit, durin42, martinvonz, #hg-reviewers Cc: mercurial-devel _______________________________________________ Mercurial-devel mailing list Mercurial-devel@mercurial-scm.org https://www.mercurial-scm.org/mailman/listinfo/mercurial-devel
