D7098: widening: trust user to give full "known" set

[martinvonz (Martin von Zweigbergk)]

martinvonz created this revision.
Herald added a reviewer: durin42.
Herald added a subscriber: mercurial-devel.
Herald added a reviewer: hg-reviewers.

REVISION SUMMARY
  The new narrow_widen wire protocol command is supposed to return data
  for exactly the revisions that the client requested (in order for it
  to be strip-free). So we should not add ancestors the client didn't
  ask for.

REPOSITORY
  rHG Mercurial

REVISION DETAIL
  https://phab.mercurial-scm.org/D7098

AFFECTED FILES
  hgext/narrow/narrowbundle2.py

CHANGE DETAILS

diff --git a/hgext/narrow/narrowbundle2.py b/hgext/narrow/narrowbundle2.py
--- a/hgext/narrow/narrowbundle2.py
+++ b/hgext/narrow/narrowbundle2.py
@@ -143,14 +143,12 @@
     # c) goto a
     #
     # until they've built up the full new state.
-    # Convert to revnums and intersect with "common". The client should
-    # have made it a subset of "common" already, but let's be safe.
-    known = set(repo.revs(b"%ln & ::%ln", known, common))
+    knownrevs = {repo.changelog.rev(n) for n in known}
     # TODO: we could send only roots() of this set, and the
     # list of nodes in common, and the client could work out
     # what to strip, instead of us explicitly sending every
     # single node.
-    deadrevs = known
+    deadrevs = knownrevs
 
     def genkills():
         for r in deadrevs:
@@ -160,7 +158,7 @@
 
     bundler.newpart(_CHANGESPECPART, data=genkills())
     newvisit, newfull, newellipsis = exchange._computeellipsis(
-        repo, set(), common, known, newmatch
+        repo, set(), common, knownrevs, newmatch
     )
     if newvisit:
         packer = changegroup.getbundler(



To: martinvonz, durin42, #hg-reviewers
Cc: mercurial-devel