Martin von Zweigbergk <martinvonz@google.com> [Thu, 07 Feb 2019 09:40:37 -0800] rev 41626
subrepo: adjust subrepo prefix before calling subrepo.removefiles() (API)
That's what we do with the matcher so it seems more consistent.
Differential Revision: https://phab.mercurial-scm.org/D5882
Martin von Zweigbergk <martinvonz@google.com> [Thu, 07 Feb 2019 09:46:36 -0800] rev 41625
subrepo: avoid calculating subrepo prefix twice for cat() (API)
Differential Revision: https://phab.mercurial-scm.org/D5881
Pulkit Goyal <pulkit@yandex-team.ru> [Thu, 07 Feb 2019 16:43:42 +0300] rev 41624
py3: make sure __repr__ returns str
No test fails but I found it while debugging test-commit-interactive-curses.t
failure.
# skip-blame because just r'' prefix
Differential Revision: https://phab.mercurial-scm.org/D5878
Pulkit Goyal <pulkit@yandex-team.ru> [Thu, 07 Feb 2019 16:44:43 +0300] rev 41623
py3: use bytes.startswith() instead of comparing with bytes[0]
This is because bytes[0] will return the ascii value and comparison will fail.
This makes test-commit-interactive-curses.t pass on Python 3.
Differential Revision: https://phab.mercurial-scm.org/D5879
Martin von Zweigbergk <martinvonz@google.com> [Fri, 08 Feb 2019 22:57:26 -0800] rev 41622
match: remove unused "exact" argument (API)
It seems that all callers use match.exact() instead of match.match(exact=True).
Differential Revision: https://phab.mercurial-scm.org/D5919
Martin von Zweigbergk <martinvonz@google.com> [Fri, 08 Feb 2019 23:03:42 -0800] rev 41621
tests: replace match.match(exact=True) by match.exact()
Differential Revision: https://phab.mercurial-scm.org/D5918
Martin von Zweigbergk <martinvonz@google.com> [Wed, 06 Feb 2019 22:52:49 -0800] rev 41620
patch: accept second matcher that applies only to copy sources (API)
See previous patch for motivation.
Differential Revision: https://phab.mercurial-scm.org/D5893
Martin von Zweigbergk <martinvonz@google.com> [Wed, 06 Feb 2019 17:46:20 -0800] rev 41619
patch: let caller pass in root-filtering matcher (API)
The --root option to `hg diff` does two things:
* Shows paths relative to the given root
* Filters paths by the given root, including copy sources
The root argument is passed through down to patch.diff(). I feel like
we can make patch.diff() more generic by not passing down the root
argument, but instead pass:
* A function for taking a repo-relative path and printing it. I want
to reuse this for showing cwd-relative paths later. This is the
actual motivation for this patch.
* A matcher that's already been filtered by the root argument
* A second matcher that filters the copy sources
This is one step towards that.
Differential Revision: https://phab.mercurial-scm.org/D5892
Martin von Zweigbergk <martinvonz@google.com> [Wed, 06 Feb 2019 17:27:43 -0800] rev 41618
patch: pass in context objects into diffhunks() (API)
It's a pretty low-level function and having the contexts in
patch.diff() makes future patches easier.
Differential Revision: https://phab.mercurial-scm.org/D5891
Nikita Slyusarev <nslus@yandex-team.com> [Thu, 07 Feb 2019 21:48:50 +0300] rev 41617
convert: handle exec bit removal while converting to svn
Subversion `putcommit` method checks original file's executablity to decide
if executable property should be removed from svn. It is checked right after
writing file contents.
Content writing is implemented using `vfs.write` and vfs seems to remove exec
bit, at least in some cases. This leads to executability checks being
ineffective. If cset contains only this ignored exec bit removal, conversion
stops with an error, because it fails to to compose svn commit properly.
This fix moves exec bit checking so that it's performed before dumping file
contents.
Added test to check executable bit removal.