gitpatch objects emitted by iterhunks() were referencing file paths unmodified
from the input patch. _applydif() made them usable by modifying the gitpatch
objects in-place with specified path strip level. The same modified objects
were then reused by iterhunks() generator. _applydiff() now copies and update
the paths which completely decouples both routines.
As a side effect, the "git" event now receives only metadata about
copies/renames to perform the necessary copies ahead of time. Other actions are
handled in the "file" event.
Patch changes are emitted by iterhunks() in two separate events: 'file' when
hunks have to be applied and 'git' to describe other modifications like copies
or mode changes. Note that a file which mode is changed and which content is
modified by the same patch will be emitted in both events. It is more
convenient to handle all file modifications in a single event. This patch
"zips" git actions with regular changes so both kinds can be emitted at the
same place.
git afile/bfile are extracted twice, once when reading a 'diff --git', and
again when reading a unified hunk. The problem is not all git blocks have
unified hunks (renames just have metadata) and they were not extracted the same
way. This is what this patch unifies.
gitpatch objects emitted by iterhunks() are modified in place by applydiff().
Processing them earlier improves iterhunks() isolation. applydiff() modifying
them should still be fixed though.
_updatedir() is no longer used by internalpatch()
The change in test-mq-missingfiles.t comes from workingbackend not considering
the missing 'b' file as changed, thus not calling addremove() on it.
This introduces a performance regression for large files, as they will be
copied just to be clobbered afterwards since binary patching does not use
deltas. But it simplifies the code and the previous optimization will be
reintroduced later in a better way.
_applydiff() patcher argument was added to help hgsubversion like extension
monkeypatching the patching process. While it could be removed at this point, I
prefer to leave it until patch.py is completely refactored and there is a valid
and tested alternative.
Most filesystem calls are already isolated in patchfile but this is not enough:
renames are performed before patchfile is available and some chmod calls are
even done outside of the applydiff call. Once all these calls are extracted
into a backend class, we can provide cleaner APIs to write to a working
directory context directly into the repository.
The local err variable would be bound to PatchError thrown and it
keeps this value even after the except block is executed.
The whole thing worked anyway since the rejected variable is set in
the except block and this makes the function return -1 when a
PatchError is thrown.
Why?
- Mercurial internal patcher works correctly for regular patches and git
patches, is much faster at least on Windows and is more extensible.
- In theory, the external patcher can be used to handle exotic patch formats. I
do not know any and have not heard about any such use in years.
- Most patch programs cannot handle git format patches, which makes the API
caller to decide either to ignore ui.patch by calling patch.internalpatch()
directly, or take the risk of random failures with valid inputs.
- One thing a patch program could do Mercurial patcher cannot is applying with
--reverse. Apparently several shelve like extensions try to use that,
including passing the "reverse" option to Mercurial patcher, which has been
removed mid-2009. I never heard anybody complain about that, and would prefer
reimplementing it anyway.
And from the technical perspective:
- The external patcher makes everything harder to maintain and implement. EOL
normalization is not implemented, and I would bet file renames, if supported
by the patcher, are not correctly recorded in the dirstate.
- No tests.
How?
- Remove related documentation
- Clearly mark patch.externalpatch() as private
- Remove the debuginstall check. This deprecation request was actually
triggered by this last point. debuginstall is the only piece of code patching
without a repository. When migrating to an integrated patch() + updatedir()
call, this was really a showstopper, all workarounds were either ugly or
uselessly complicated to implement. If we do not support external patcher
anymore, the debuginstall check is not useful anymore.
- Remove patch.externalpatch() after 1.9 release.
The patch changes the output of "hg diff --stat" when one file whose filename
has spaces has changed, making it get the full filename instead of just the
substring between the last space and the end of the filename.
It also changes the diffstat generated by "hg email -d" when one of the commit
messages starts with "diff". Because of the regex used to parse the filename,
the diffstat generated by "hg email -d" will still be not correct if a commit
message starts with "diff -r ".
Before the patch Mercurial has the following behavior:
$ echo "foobar">"file with spaces"
$ hg add "file with spaces"
$ hg diff --stat
spaces | 1 +
1 files changed, 1 insertions(+), 0 deletions(-)
$ hg diff --git --stat
file with spaces | 1 +
1 files changed, 1 insertions(+), 0 deletions(-)
After the patch:
$ echo "foobar">"file with spaces"
$ hg add "file with spaces"
$ hg diff --stat
file with spaces | 1 +
1 files changed, 1 insertions(+), 0 deletions(-)
$ hg diff --git --stat
file with spaces | 1 +
1 files changed, 1 insertions(+), 0 deletions(-)
Before the patch:
$ hg add mercurial/patch.py tests/tests-diffstat.t
$ hg commit -m "diffstat: fix parsing of filenames"
$ hg email -d --test tip
This patch series consists of 1 patches.
diffstat: fix parsing of filenames
[...]
filenames | 0
mercurial/patch.py | 6 ++++--
tests/test-diffstat.t | 17 +++++++++++++++++
3 files changed, 21 insertions(+), 2 deletions(-)
[...]
After the patch:
$ hg email -d --test tip
This patch series consists of 1 patches.
diffstat: fix parsing of filenames
[...]
mercurial/patch.py | 6 ++++--
tests/test-diffstat.t | 17 +++++++++++++++++
3 files changed, 21 insertions(+), 2 deletions(-)
[...]
Do not pass reject file content to patchfile.writelines() to:
- Avoid line endings transformations
- Avoid polluting overriding implementations with unrelated data. They should
override write_rej() to deal or ignore reject files properly.
Bug report, analysis and original patch and test by
Shun-ichi GOTO <shunichi.goto@gmail.com>