view tests/test-push-checkheads-unpushed-D1.t @ 5572:d7d954b6d518

evolve: remove swapping nodes logic while resolving divergence

Before this patch, while solving content-divergence if one of the
divergent cset need to relocate to resolution parent we used to set that
cset to "other" side; so it was like we are always going to relocate "other"
side which made us to swap the two if "divergent" cset need to relocate.

Above mentioned solution for relocating cset was not good and led us to incorrect
output in some cases.

So this patch, simplify the logic and remove the swapping thing and clearly depend
on resolution parent to check if any of the divergent cset needs to relocate.
This also means that now in some cases if both the divergent csets needs
to relocate, both will be relocated. Even before this patch, there are some cases
where both the divergent csets relocated but not in every case where it should
have.

Coming to the changes in test files:
1) changes in order of merging the two divergent csets is due to fact that
   now "divergent" and "other" are not swapped and hence "divergent" will be at
   local side and "other" will be at "other", which also results in change in
   the hash id for some cases.

2) extra changesets: which is the result of the fact that now we don't miss
   any case where both the divergent csets should be based on resolution parent
   before performing the merge -> which caught some cases and performed relocation,
   hence extra changesets. Although you can find that there is no changes in file
   content because of these relocation, but I remember loosing some data because
   of not relocating the "divergent", so why leave any risk out there by not
   relocating "other" to the resolution parent.

====================================
Testing head checking code: Case D-1
====================================

Mercurial checks for the introduction of new heads on push. Evolution comes
into play to detect if existing branches on the server are being replaced by
some of the new one we push.

This case is part of a series of tests checking this behavior.

Category D: remote head is "obs-affected" locally, but result is not part of the push
TestCase 1: remote head is rewritten, but successors is not part of the push

.. old-state:
..
.. * 1 changeset branch
..
.. new-state:
..
.. * 1 changeset branch succeeding the old branch
.. * 1 new unrelated branch
..
.. expected-result:
..
.. * pushing only the unrelated branch: denied
..
.. graph-summary:
..
..   A ø⇠○ A'
..     |/
..     | ◔ B
..     |/
..     ●

  $ . $TESTDIR/testlib/push-checkheads-util.sh

Test setup
----------

  $ mkdir D1
  $ cd D1
  $ setuprepos
  creating basic server and client repo
  updating to branch default
  2 files updated, 0 files merged, 0 files removed, 0 files unresolved
  $ cd client
  $ hg up 0
  0 files updated, 0 files merged, 1 files removed, 0 files unresolved
  $ mkcommit A1
  created new head
  $ hg debugobsolete `getid "desc(A0)" ` `getid "desc(A1)"`
  1 new obsolescence markers
  obsoleted 1 changesets
  $ hg up 0
  0 files updated, 0 files merged, 1 files removed, 0 files unresolved
  $ mkcommit B0
  created new head
  $ hg log -G --hidden
  @  74ff5441d343 (draft): B0
  |
  | o  f6082bc4ffef (draft): A1
  |/
  | x  8aaa48160adc (draft): A0
  |/
  o  1e4be0697311 (public): root
  

Actual testing
--------------

  $ hg push -r 'desc(B0)'
  pushing to $TESTTMP/D1/server (glob)
  searching for changes
  abort: push creates new remote head 74ff5441d343!
  (merge or see 'hg help push' for details about pushing new heads)
  [255]

  $ cd ../..