| Age | Commit message (Collapse) | Author |
|---|
|
|
|
|
|
|
|
A diff may contain just headers and no hunk. When washing a diff
directly, we already detected that by checking whether we have reached
the beginning of the next diff or eob.
When the diff is part of a log, we additionally have to also check
whether we have reached the next log heading.
Alternatively we could pass along the limit from `magit-log-wash-rev'
(as a marker, because we modify the buffer text), but that would
complicate matters.
Closes #5454.
|
|
As for `completing-read-multiple', if REQUIRE-MATCH is t, then all
inputs must appear in the collection, but it is again possible to
select nothing.
Forge's uses of this function need this behavior. In [1: 4e2dd7ba6f]
I should only have added the new `any' value, instead of also changing
the meaning of t.
Closes magit/forge#814.
1: 2025-08-16 4e2dd7ba6f54b2251a702b4c0416467d7315407a
magit-completing-read-multiple: Support requiring non-empty input
|
|
|
|
Generally, when we bind SPC to some "scroll down" command, we also
bind DEL *and* S-SPC to the respective "scroll up" command. Do the
same here.
|
|
Before the menu `magit-revision-jump' was added, when point is on a
file in a diff "j" (`magit-jump-to-diffstat-or-diff') jumped to the
respective file within the diffstat. For `magit-diff-mode' it still
does that.
In `magit-revision-mode' "j" is now bound to `magit-revision-jump',
with partially shadows the above behavior. To continue to allow
quickly jumping between a file in a revision diff and the same file
listed in the diffstat, `magit-revision-jump' pivots to calling
`magit-jump-to-diffstat-or-diff' when invoked in those locations.
The hope was that would preserve enough of the old behavior to make
everyone happy. I.e., on a file jump to the other place that file
is listed (as before), everywhere else (or with a prefix argument)
show the menu, letting the user explicitly choose where to jump to
this time around (previously this would have jumped to the diffstat).
Users who want the old behavior back can do:
(keymap-unset magit-revision-mode-map "j" t)
Instead of adding a more explicit knob to restore the old behavior,
this commit improves a variation:
If `magit-jump-revision-diffstat' is invoked on a hunk (via "j s"),
that now goes to the respective file in the diffstat, instead of to
the heading of the diffstat itself.
The hope is that going to a more precise location outweighs the added
cost of having to press "j s" instead of just "j" to go to the diffstat.
Do the same for `magit-jump-to-revision-diffstat', so that those users,
who remove the new binding in `magit-revision-mode-map', can benefit
too.
Now let's just hope nobody dislikes going to the file heading instead of
to the diffstat heading.
Closes #5442.
|
|
The next commit will make changes that `magit-define-section-jumper'
does not support.
|
|
As for related commands.
|
|
|
|
|
|
If the user has set `enable-local-variables' to nil, we want to
respect that. `normal-mode's FIND-FILE argument wasn't designed for
our use case, so we have to use this strange invocation to achieve
that.
Unfortunately, it is not possible to prevent every non-nil value being
interpreted as t. In the future we might re-implement `normal-mode'
to respect those values as well.
|
|
Closes #5451.
|
|
|
|
Closes #5450.
|
|
|
|
|
|
|
|
|
|
If the SECTION:S argument is a single section, pass that along to
the APPLY function again. Regression added in [1: d3f65e92fc].
Closes #5446.
1: 2025-09-18 d3f65e92fc83d9d63d947eb95c64b372c28abb03
magit-discard-apply: Absorb magit-discard-apply
|
|
Clarify the docstrings of `magit-section-hidden'
and `magit-section-hidden-body'.
|
|
|
|
Here "primus" is short for "primus inter pares".
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
The idea was to only ask `package' about what Magit version it
installed when it is actually being used to manage packages.
Loading it unconditionally goes counter to those intentions.
Also, `straight' makes noise if it notices that `package' has
been loaded as well.
Closes #5443.
Closes #5445.
|
|
We require `vc-git' for some minor functionality. Loading that took
care of making some of our own dependencies available. Emacs commit
[1: 2ee74aca554] stops requiring `message', and as a result we now
have to require that and `dired' and `epa' in the appropriate places
ourselves.
1: 2025-09-12 2ee74aca554786c20193b528363eb426e4f7f65c
Delete obsolete log-edit-beginning-of-line
|
|
Add it now on "main", without actually using it, to make it easier
to switch between "main" and the feature branch, which does use it.
|
|
|
|
|
|
|
|
Closes #5432.
|
|
|
|
|
|
|
|
Closes #5437.
|
|
|
|
|
|
|
|
Closes #5436.
|
|
|
|
That is how I always order such pairs. Not sure what happened here.
|
|
|
|
|
