Pierre-Yves David <pierre-yves.david@octobus.net> [Mon, 14 Dec 2020 14:31:15 +0100] rev 46138
test-narrow: mitigate the flakyness of test-narrow-shallow
The actual bug still need fixing, but in the meantime the flakyness is wasting
everybody's time.
Differential Revision: https://phab.mercurial-scm.org/D9593
Matt Harbison <matt_harbison@yahoo.com> [Mon, 14 Dec 2020 18:03:29 -0500] rev 46137
hook: set `HGPLAIN=1` for external hooks
External hooks will probably turn around and run hg commands, so this seems like
a reasonable convenience for people who miss it in the documentation.
There are no test changes because `printenv.py` filters out anything without a
"HG_" prefix. It wouldn't be a useful test anyway, because this is already
inherited from the test environment.
Differential Revision: https://phab.mercurial-scm.org/D9605
Matt Harbison <matt_harbison@yahoo.com> [Mon, 14 Dec 2020 17:59:10 -0500] rev 46136
setup: exclude the git extension from py2 builds
This can't be built on Windows with the py2 compiler, and while old versions can
be installed via pip on Linux, I can't get the tests to run (even with py3.8)
using pygit2 0.28.2. Some manually run commands work, and others spew stack
traces that don't occur with the current 1.4.0 release using py3.
Differential Revision: https://phab.mercurial-scm.org/D9604
Simon Sapin <simon.sapin@octobus.net> [Mon, 14 Dec 2020 14:59:23 +0100] rev 46135
rust: replace most "operation" structs with functions
The hg-core crate has a partially-formed concept of "operation",
represented as structs with constructors and a `run` method.
Each struct’s contructor takes different parameters,
and each `run` has a different return type.
Constructors typically don’t do much more than store parameters
for `run` to access them.
There was a comment about adding an `Operation` trait
when the language supports expressing something so general,
but it’s hard to imagine how operations with such different APIs
could be used in a generic context.
This commit starts removing the concept of "operation",
since those are pretty much just functions.
Differential Revision: https://phab.mercurial-scm.org/D9595
Simon Sapin <simon.sapin@octobus.net> [Mon, 14 Dec 2020 13:47:44 +0100] rev 46134
rust: change &PathBuf parameters to &Path
This is just as useful in function bodies, and a less strict requirement for callers.
Differential Revision: https://phab.mercurial-scm.org/D9594
Martin von Zweigbergk <martinvonz@google.com> [Wed, 16 Dec 2020 21:06:29 -0800] rev 46133
extdiff: add --from/--to and deprecate -r, as was done for `hg diff`
I recently replaced `hg diff`'s `-r` flag by `--from` and `--to` and
then deprecated the `-r` flag. This patch repeats that for
`hg extdiff`.
Differential Revision: https://phab.mercurial-scm.org/D9624
Martin von Zweigbergk <martinvonz@google.com> [Wed, 16 Dec 2020 22:30:07 -0800] rev 46132
extdiff: fix crash when showing diff from wdir()
Differential Revision: https://phab.mercurial-scm.org/D9623
Martin von Zweigbergk <martinvonz@google.com> [Wed, 16 Dec 2020 22:16:05 -0800] rev 46131
extdiff: pass contexts instead of nodeids into diffrevs()
This just avoids some unnecessary lookups.
Differential Revision: https://phab.mercurial-scm.org/D9622
Martin von Zweigbergk <martinvonz@google.com> [Wed, 16 Dec 2020 21:40:27 -0800] rev 46130
tests: remove undefined (empty) $opt from test-extdiff.t
It's been undefined for a long time (at least since .t unification).
Differential Revision: https://phab.mercurial-scm.org/D9621
Pierre-Yves David <pierre-yves.david@octobus.net> [Tue, 21 Apr 2020 15:13:13 +0200] rev 46129
copies-rust: pre-introduce a PathToken type and use it where applicable
Handling string all the time is quite slow. So we are about to use more
efficient "token". We do some of the churn early to clarify the next changesets.
Differential Revision: https://phab.mercurial-scm.org/D9492