This is an archive of the discontinued Mercurial Phabricator instance.

Differential D8727

rebase: add test case that shows that inmemory rebase does not preserve merges
ClosedPublic

Authored by mjacob on Jul 10 2020, 6:54 PM.

Details

Reviewers
None
Group Reviewers
hg-reviewers
Commits
rHGab33782deb84: rebase: add test case that shows that inmemory rebase does not preserve merges

Diff Detail

Repository
rHG Mercurial
Lint
Automatic diff as part of commit; lint not applicable.
Unit
Automatic diff as part of commit; unit tests not applicable.

Event Timeline

mjacob created this revision.Jul 10 2020, 6:54 PM
martinvonz added a subscriber: martinvonz.Jul 13 2020, 5:01 PM

This patch is based on c4e02bd172d57879717f8bc04c4b5bbc9268e36d, but I don't see any such commit. Can you rebase it to a public commit?

mjacob added a comment.Jul 13 2020, 5:12 PM
In D8727#130344, @martinvonz wrote:

This patch is based on c4e02bd172d57879717f8bc04c4b5bbc9268e36d, but I don't see any such commit. Can you rebase it to a public commit?

I rebased D8724 to D8733 onto @. Strictly speaking, D8727 is not dependent on D8725 or D8724, but D8729 is dependent on D8724, so it made sense to keep all of them as a linear chain.

martinvonz added a comment.Jul 13 2020, 5:15 PM
In D8727#130348, @mjacob wrote:
In D8727#130344, @martinvonz wrote:

This patch is based on c4e02bd172d57879717f8bc04c4b5bbc9268e36d, but I don't see any such commit. Can you rebase it to a public commit?

I rebased D8724 to D8733 onto @. Strictly speaking, D8727 is not dependent on D8725 or D8724, but D8729 is dependent on D8724, so it made sense to keep all of them as a linear chain.

It does seem to be textually dependent on D8724 -- the patch didn't apply before. Thanks!

Revision Contents
Changeset List

PathPackages
Mtests/test-rebase-inmemory.t (47 lines)

Diff 21879

tests/test-rebase-inmemory.t

o 3: bye from foo o 3: bye from foo
| foo | foo
| @ 2: some unrelated changes | @ 2: some unrelated changes
| | some_other_file | | some_other_file
o | 1: hi from foo o | 1: hi from foo
|/ foo |/ foo
o 0: r0 o 0: r0
r0 r0
$ cd ..
Changesets that become empty should not be committed. Merges are not empty by
definition.
$ hg init keep_merge
$ cd keep_merge
$ echo base > base; hg add base; hg ci -m base
$ echo test > test; hg add test; hg ci -m a
$ hg up 0 -q
$ echo test > test; hg add test; hg ci -m b -q
$ hg up 0 -q
$ echo test > test; hg add test; hg ci -m c -q
$ hg up 1 -q
$ hg merge 2 -q
$ hg ci -m merge
$ hg up null -q
$ hg tglog
o 4: 59c8292117b1 'merge'
|\
| | o 3: 531f80391e4a 'c'
| | |
| o | 2: 0194f1db184a 'b'
| |/
o / 1: 6f252845ea45 'a'
|/
o 0: d20a80d4def3 'base'
FIXME: It's broken for inmemory merges.
$ hg rebase -s 2 -d 3
rebasing 2:0194f1db184a "b"
note: not rebasing 2:0194f1db184a "b", its destination already has all its changes
rebasing 4:59c8292117b1 "merge" (tip)
note: not rebasing 4:59c8292117b1 "merge" (tip), its destination already has all its changes (true !)
saved backup bundle to $TESTTMP/keep_merge/.hg/strip-backup/0194f1db184a-aee31d03-rebase.hg
#if false
$ hg tglog
o 3: 506e2454484b 'merge'
|\
| o 2: 531f80391e4a 'c'
| |
o | 1: 6f252845ea45 'a'
|/
o 0: d20a80d4def3 'base'
#endif