[Bug 4210] New: Rebasing when there are unstable changesets leads to divergent changesets

[mercurial-bugs at selenic.com]

http://bz.selenic.com/show_bug.cgi?id=4210

          Priority: normal
            Bug ID: 4210
                CC: mercurial-devel at selenic.com
          Assignee: bugzilla at selenic.com
           Summary: Rebasing when there are unstable changesets leads to
                    divergent changesets
          Severity: bug
    Classification: Unclassified
                OS: All
          Reporter: gregory.szorc at gmail.com
          Hardware: All
            Status: UNCONFIRMED
           Version: unspecified
         Component: rebase
           Product: Mercurial

I found a user experience with 2.9.1+212-8a6a86c9a5b5 + mutable-history
0ea9cecbbd28 (a few changesets off tip) that is quite unpleasant. I'm not sure
if it is a bug or by design. I'm not sure if it is a bug with mutable-history
or core, so filing here.

I have a repo that looks like this:

A --> B --> C --> D --> E

I amend a commit to produce some unstable changesets:

$ hg up C
<make some changes>
$ hg amend
2 new unstable changesets

Cool. Now, if I had my wits about me, I'd run `hg evolve -a` right now. But I
forgot. Instead, I do a pull:

$ hg pull

Now my history is like:

A --> B --> C --> D --> E
 \      \
  \      \- C'
   \
    F -> G

. is C' (rebased C). D and E are unstable.

I want to rebase my feature head on top of the newly pulled head.

$ hg rebase -d G

Mercurial churns for a while and I immediately notice something weird: the
progress bars indicate it's rebasing 3 patches. Not just C', but D and E as
well. I think "ok, it's being smart and evolving for me!" But then rebase
finishes:

2 new divergent changesets

Oh, hmm. That's weird. 

Well, I want to push the rebased C' to the remote, so:

$ hg out -r .
<it shows something that looks sane>

$ hg push -r .
abort: push includes divergent changeset: a978f03b56fd!
(use 'hg evolve' to get a stable history or --force to ignore warnings)

And this is the point where I start breaking out `hg --hidden log` to try to
figure out WTF happened to my repo.

With much trepidation, I eventually run `hg evolve -a` and it seems to do the
right thing. Although, the output is very unsettling:

merge:[201556] Bug 985084 - Experiment add-ons should be disabled by default;
r=Unfocused
with: [201568] Bug 985084 - Experiment add-ons should be disabled by default;
r=Unfocused
base: [201473] Bug 985084 - Experiment add-ons should be disabled by default;
r=Unfocused
updating to "local" conflict
4 files updated, 0 files merged, 0 files removed, 0 files unresolved
1 files updated, 0 files merged, 0 files removed, 0 files unresolved
11 new unstable changesets

"updating to 'local' conflict" - that sounds scary!

I start looking at `hg diff -c` and everything looks sane. So, I eventually
arrive at the conclusion that everything is fine.

But the whole experience has me scratching my head.

Why was . not updated to a stable changeset after the rebase?

Why did the rebase include unstable changesets (C'::) and create the divergent
changesets in the first place?

I'm pretty confident there's a bug in here somewhere. My apologies if this
should be filed in evolution's bug tracker.

-- 
You are receiving this mail because:
You are on the CC list for the bug.