Patchwork [V2] changegroup: fix file linkrevs during reorders (issue4462)

login
register
mail settings
Submitter Durham Goode
Date Nov. 21, 2014, 9:08 p.m.
Message ID <b5e62dcd56b5454f902e.1416604109@dev2000.prn2.facebook.com>
Download mbox | patch
Permalink /patch/6820/
State Accepted
Commit cc0ff93d0c0c29526465a1d07b9eddf5b2da51dd
Headers show

Comments

Durham Goode - Nov. 21, 2014, 9:08 p.m.
# HG changeset patch
# User Durham Goode <durham@fb.com>
# Date 1416529857 28800
#      Thu Nov 20 16:30:57 2014 -0800
# Node ID b5e62dcd56b5454f902e8a7080fcd5a7c9149535
# Parent  a179db3db9b96b38c10c491e6e7e7ad5f40a7787
changegroup: fix file linkrevs during reorders (issue4462)

Previously, if reorder was true during the creation of a changegroup bundle,
it was possible that the manifest and filelogs would be reordered such that the
resulting bundle filelog had a linkrev that pointed to a commit that was not
the earliest instance of the filelog revision. For example:

With commits:

0<-1<---3<-4
  \       /
   --2<---

if 2 and 3 added the same version of a file, if the manifests of 2 and 3 have
their order reversed, but the changelog did not, it could produce a filelog with
linkrevs 0<-3 instead of 0<-2, which meant if commit 3 was stripped, it would
delete that file data from the repository and commit 2 would be corrupt (as
would any future pulls that tried to build upon that version of the file).

The fix is to make the linkrev fixup smarter. Previously it considered the first
manifest that added a file to be the first commit that added that file, which is
not true. Now, for every file revision we add to the bundle we make sure we
attach it to the earliest applicable linkrev.

I can only repro this issue in a large internal repo, so I don't have a good
test case to add a test for (since it depends on the exact commit vs. manifest
vs. filelog orders).
Matt Mackall - Nov. 22, 2014, 1:24 a.m.
On Fri, 2014-11-21 at 13:08 -0800, Durham Goode wrote:
> # HG changeset patch
> # User Durham Goode <durham@fb.com>
> # Date 1416529857 28800
> #      Thu Nov 20 16:30:57 2014 -0800
> # Node ID b5e62dcd56b5454f902e8a7080fcd5a7c9149535
> # Parent  a179db3db9b96b38c10c491e6e7e7ad5f40a7787
> changegroup: fix file linkrevs during reorders (issue4462)

Seems to want to be on stable and applies there, so on stable it goes. I
guess I'll drop the "I don't have a test" comment as well. Thanks!

Patch

diff --git a/mercurial/changegroup.py b/mercurial/changegroup.py
--- a/mercurial/changegroup.py
+++ b/mercurial/changegroup.py
@@ -325,6 +325,7 @@  class cg1packer(object):
         # for progress output
         msgbundling = _('bundling')
 
+        clrevorder = {}
         mfs = {} # needed manifests
         fnodes = {} # needed file nodes
         changedfiles = set()
@@ -334,6 +335,7 @@  class cg1packer(object):
         # Returns the linkrev node (identity in the changelog case).
         def lookupcl(x):
             c = cl.read(x)
+            clrevorder[x] = len(clrevorder)
             changedfiles.update(c[3])
             # record the first changeset introducing this manifest version
             mfs.setdefault(c[0], x)
@@ -349,13 +351,16 @@  class cg1packer(object):
         # Returns the linkrev node (collected in lookupcl).
         def lookupmf(x):
             clnode = mfs[x]
-            if not fastpathlinkrev:
+            if not fastpathlinkrev or reorder:
                 mdata = mf.readfast(x)
                 for f, n in mdata.iteritems():
                     if f in changedfiles:
                         # record the first changeset introducing this filelog
                         # version
-                        fnodes.setdefault(f, {}).setdefault(n, clnode)
+                        fclnodes = fnodes.setdefault(f, {})
+                        fclnode = fclnodes.setdefault(n, clnode)
+                        if clrevorder[clnode] < clrevorder[fclnode]:
+                            fclnodes[n] = clnode
             return clnode
 
         mfnodes = self.prune(mf, mfs, commonrevs, source)
@@ -368,7 +373,7 @@  class cg1packer(object):
         needed = set(cl.rev(x) for x in clnodes)
 
         def linknodes(filerevlog, fname):
-            if fastpathlinkrev:
+            if fastpathlinkrev and not reorder:
                 llr = filerevlog.linkrev
                 def genfilenodes():
                     for r in filerevlog:
diff --git a/tests/test-generaldelta.t b/tests/test-generaldelta.t
old mode 100755
new mode 100644
--- a/tests/test-generaldelta.t
+++ b/tests/test-generaldelta.t
@@ -22,3 +22,50 @@  inserted due to big distance from its pa
   >>> gdsize = os.stat("gdrepo/.hg/store/00manifest.i").st_size
   >>> if regsize < gdsize:
   ...     print 'generaldata increased size of manifest'
+
+Verify rev reordering doesnt create invalid bundles (issue4462)
+This requires a commit tree that when pulled will reorder manifest revs such
+that the second manifest to create a file rev will be ordered before the first
+manifest to create that file rev. We also need to do a partial pull to ensure
+reordering happens. At the end we verify the linkrev points at the earliest
+commit.
+
+  $ hg init server --config format.generaldelta=True
+  $ cd server
+  $ touch a
+  $ hg commit -Aqm a
+  $ echo x > x
+  $ echo y > y
+  $ hg commit -Aqm xy
+  $ hg up -q '.^'
+  $ echo x > x
+  $ echo z > z
+  $ hg commit -Aqm xz
+  $ hg up -q 1
+  $ echo b > b
+  $ hg commit -Aqm b
+  $ hg merge -q 2
+  $ hg commit -Aqm merge
+  $ echo c > c
+  $ hg commit -Aqm c
+  $ hg log -G -T '{rev} {shortest(node)} {desc}'
+  @  5 ebb8 c
+  |
+  o    4 baf7 merge
+  |\
+  | o  3 a129 b
+  | |
+  o |  2 958c xz
+  | |
+  | o  1 f00c xy
+  |/
+  o  0 3903 a
+  
+  $ cd ..
+  $ hg init client
+  $ cd client
+  $ hg pull -q ../server -r 4
+  $ hg debugindex x
+     rev    offset  length   base linkrev nodeid       p1           p2
+       0         0       3      0       1 1406e7411862 000000000000 000000000000
+