Martin von Zweigbergk <martinvonz@google.com> [Tue, 25 Jun 2019 14:25:03 -0700] rev 42594
copies: filter invalid copies only at end of pathcopies() (issue6163)
copies._filter() filters out copies whose source file does not exist
in the start commit or whose target file does not exist in the end
commit. We do that after chaining copies with dirstate copies or
backward renames from another branch. We also do at the end of the
changeset-centric copy tracing. The filtering means that we will
remove copies to/from files that did not exist in some intermediate
commit. That is inconsistent with what we do if a file has been
deleted and then re-added (we allow updating across that).
Copying the two first examples from issue6163:
@ 4 'rename x to y'
|
o 3 'add x again'
|
o 2 'remove x'
|
| o 1 'modify x'
|/
o 0 'add x'
@ 4 'rename x to y'
|
o 3 'add x again'
|
| o 2 'modify x'
| |
| o 1 'add x'
|/
o 0 'base'
When doing `hg rebase -r 1 -d 4` in the first case, it succeeds, but
`hg rebase -r 2 -d 4` in the second case does not. That's because we
chain and filter via commit 0, which does not have file 'x' in the
second case. IMO, that's clearly inconsistent. So this patch removes
the filtering step so it only happens at the end. If a file was
temporarily removed, whether via a merge base or not, it will now
still be considered the same file. That fixes issue6163 for the
changeset-centric case.
Differential Revision: https://phab.mercurial-scm.org/D6603
Martin von Zweigbergk <martinvonz@google.com> [Tue, 25 Jun 2019 13:46:55 -0700] rev 42593
copies: inline _chainandfilter() to prepare for next patch
Differential Revision: https://phab.mercurial-scm.org/D6602
Martin von Zweigbergk <martinvonz@google.com> [Tue, 25 Jun 2019 13:33:49 -0700] rev 42592
copies: remove most early returns from pathcopies() and _forwardcopies()
I want to split up _chainandfilter() more so the call to _filter()
consistently happens at the end of pathcopies(). This prepares for
that change.
Differential Revision: https://phab.mercurial-scm.org/D6601
Martin von Zweigbergk <martinvonz@google.com> [Fri, 28 Jun 2019 09:01:45 -0700] rev 42591
copies: move short-circuiting of dirstate copies out of _forwardcopies()
I'd like to move the filtering of copies we do after chaining to the
end of all chaining (in a single place in pathcopies()). One problem
that came up when trying that was that we allow things like `hg cp -f
<file> <existing file>` so the user can later amend that in. Filtering
at the end would mean that we remove those copies. That would break
`hg st -C`. This patch therefore moves the short-circuiting of
dirstate copies into pathcopies() so we can more easily handle the
dirstate-only case differently.
I initially thought this might change some behavior when the user does
`hg status --rev 'wdir()' --rev .` during an uncommitted merge, since
_backwardrenames() would reverse the copies in that case. However, I
couldn't come up with a test case where it made a difference.
Differential Revision: https://phab.mercurial-scm.org/D6600
Martin von Zweigbergk <martinvonz@google.com> [Fri, 21 Jun 2019 16:59:29 -0700] rev 42590
tests: add more tests of copy tracing with removed and re-added files
We had a test where the destination of a copy was removed and then
added back. This patch adds similar cases where the break in history
instead happens to the source file. There are three versions of this:
1. The break happens before the rename.
2. The break happens on a branch parallel to the rename (where copy
tracing is done via the merge base)
3. The source is added on each side of the merge base. The break in
history is thus in the form of a deletion when going backwards to
the merge base and the re-add happens on the other branch.
I've also added calls to `hg graft` in these cases to show the
breakage in issue 6163.
Another factor in these cases is matching nodeid (checked in
copies._tracefile()). I've made two copies each of the cases to show
the impact of that. One of these is the same as a test in
test-rename-merge1.t, so I also deleted that test from there.
Some of these tests currently fail, where "fail" is based on my
current thinking of how things should work. I had initially thought
that we should be more strict about not tracing copies across commits
where the file did not exist, but issue 6163 made me reconsider.
The only test case here that behaved differently in 4.9 is the
exact case reported in issue 6163.
Differential Revision: https://phab.mercurial-scm.org/D6599
Martin von Zweigbergk <martinvonz@google.com> [Mon, 01 Jul 2019 14:24:51 -0700] rev 42589
tests: split out tests for unrelated copy source/target into separate file
I've realized only recently how many cases there are where a file is
treated differently if it's considered "related" to another file (not
deleted and re-added). I'll add more tests for some of these cases
soon.
Differential Revision: https://phab.mercurial-scm.org/D6598
Kyle Lippincott <spectral@google.com> [Mon, 24 Jun 2019 16:01:01 -0700] rev 42588
subrepos: make last line of prompts <40 english chars (issue6158)
Differential Revision: https://phab.mercurial-scm.org/D6572
Kyle Lippincott <spectral@google.com> [Mon, 24 Jun 2019 16:00:39 -0700] rev 42587
largefiles: make last line of prompts <40 english chars (issue6158)
Differential Revision: https://phab.mercurial-scm.org/D6571
Yuya Nishihara <yuya@tcha.org> [Sun, 30 Jun 2019 18:32:43 +0900] rev 42586
rust-dirstate: add helper to iterate ancestor paths
This is modeled after std::path::Path::ancestors().
find_dirs(b"") yields b"" because Mercurial's util.finddirs() works in that
way, and the test case for DirsMultiset expects such behavior.
Matt Harbison <matt_harbison@yahoo.com> [Tue, 09 Jul 2019 20:51:48 -0400] rev 42585
tests: update test-commit-interactive.t for no-execbit platforms
These changes correspond with f802a75da585.
Differential Revision: https://phab.mercurial-scm.org/D6624