| Commit message (Collapse) | Author | Age |
|
|
|
| |
Closes #85.
|
|
|
|
| |
Closes #83.
|
|
|
|
| |
Closes #81.
|
|
|
|
|
|
| |
In Emacs 28 `server-switch-buffer' gains an additional argument.
Closes #82.
|
|
|
|
|
|
|
|
| |
Callers may pass invalid keyword arguments to `make-process', which
is their problem and should not result in me having to explain that
to anyone.
Closes #79 and https://github.com/hlissner/doom-emacs/issues/3384.
|
|
|
|
|
|
|
| |
This is necessary because in Emacs 27, `make-process' can also start
remote processes.
Fixes #75.
|
| |
|
|
|
|
| |
Closes #74.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The `-e' flag of `echo' that makes the command interpret
backslash-escaped sequences (in particular "\037") is not supported
in all shells.
Shells that do not support it print the flag instead.
So here we replace `echo' by `printf' which seems to be better
supported and does the interpretation of backslash-escaped sequence
by default.
Re #71.
Closes #72.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The previous value of `with-editor-sleeping-editor' variable was using
the \x1f character in the elisp string and sending it to the
shell/term process.
For `shell-mode' this is not an issue since the character is not
interpreted by the shell. But in `term-mode' the terminal emulation
interprets it and remove previous characters, the "$0" string. So we
end up with an EDITOR environment variable that contains an echo
command with "…OPEN IN…" instead of "…OPEN $0^_ IN…" (here "^_" is
used to represent the \x1f byte).
This change fixes the issue by using `-e' flag of `echo' and using the
`\037' octal escape sequence for the \x1f byte.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
`git-commit-post-finish-hook' is intended to be set locally. Because
the buffer in which it is locally does not exist anymore by the time
we have reached the "post finish" time, we use a temporary buffer and
carry over the value from the other buffer.
However we accidentally only copied the value of the hook variable if
it already non-nil locally in the temporary buffer (which should not
be the case) or non-nil globally (which might very well not be the
case.
Closes #68.
|
| |
|
|
|
|
|
| |
Re #64.
Closes #66.
|
|
|
|
|
|
|
|
|
|
|
| |
Previously the value of `async-bytecomp-allowed-packages' always was a
list packages and the special symbol `all'. We failed to handle `all'.
Now the value can also be one of the atoms `t' and `all'.
Adjust our related code to handle all valid values.
Re https://github.com/jwiegley/emacs-async/pull/113.
Re https://github.com/magit/magit/pull/3981.
|
| |
|
|
|
|
|
| |
Modes can set that to display their own usage message instead
of the default one.
|
|
|
|
|
|
|
| |
We try to display our message after the one displayed by
`server-execute', which is run using a timer. The timer
starts "immediately" but apparently that takes longer
than 0.01 seconds, so we now wait five times that long.
|
| |
|
|
|
|
| |
Closes #64.
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
| |
Run `with-editor-post-{finish,cancel}-hook' in the same directory as
was current in the buffer from which `with-editor-{finish,cancel}'
was invoked. This is necessary because running `with-editor-return'
likely changes the value of `default-directory', since it changes
the current directory.
|
| |
|
|
|
|
| |
The default value was modified between 2.7.0 and 2.7.1.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
When `git-commit' is invoked from a sub-directory, then it changes the
working directory to the root of the working tree before invoking the
editor. Additionally it ask the editor to edit a relative path in
most cases. That path is relative to the top-level, so we need the
sleeping editor to also print the name of that directory; or we would
end up trying to edit e.g. "/path/to/repo/subdir/.git/COMMIT_EDITMSG"
instead of "/path/to/repo/.git/COMMIT_EDITMSG".
Actually, as of 2.19.0, `git-commit' no longer does this, it uses an
absolute file-name now. But programs might still do it and that is
legitimate. `$EDITOR' inherits the working directory from its parent
process, so it normally isn't a problem if the file-name is relative.
It only a problem here because `$EDITOR' in hour case forwards the
edit request to a running `emacs' instance, which is not a child
process and therefore does not inherit the working directory.
Fixes #55.
|
| |
|
|
|
|
| |
The old heading was a bit silly.
|
|
|
|
|
|
| |
When `server-process` dies, it still keeps the reference to the died
process with status `closed` and is evaluated as non-nil, therefore does
not restart correctly in such cases.
|
|
|
|
|
|
| |
The selected function are the function describe as entry point in
README.md plus with-editor-export-hg-editor that has the same role
than with-editor-export-git-editor, but for another VCS.
|
| |
|
|
|
|
|
|
|
| |
This fixes the issues described in #40 and reintroduces the issue
described in #23. Users who do not care about #40 but don't like #23
can set the new option `with-editor-shell-command-use-emacsclient' to
nil.
|
|
|
|
|
|
|
|
|
|
|
|
| |
Because `emacsclient' only passed the environment along when
its `--create-frame' argument is used, we cannot tell whether
`with-editor' is involved, and therefore should not enable
`with-editor-mode'.
If we use the sleeping editor, then we know that `with-editor'
is involved, so always use that in our advice to `shell-command'.
Fixes #23.
|
|
|
|
| |
Require pcase at compile time only.
|
| |
|
|
|
|
|
|
|
|
|
| |
... because `with-editor-process-filter' may bind it to nil.
When we still used `file-remote-p' instead of `tramp-tramp-file-p' in
`with-editor-output-filter' (i.e. before 9fc047751b53), then that did
not matter because the latter simply returns nil when its argument is
nil, while the former raises an error.
|
| |
|
|
|
|
|
| |
Previously we just reinstated to already set filter.
This completes fab38c451da40db17d021009ffbc53199b5e2237.
|
|
|
|
|
|
| |
Fixes #34.
Modified-by: Jonas Bernoulli <jonas@bernoul.li>
|
|
|
|
|
|
|
|
|
| |
Emacs usually identifies itself as "N.M.1" or "N.0.[59][0-9]"
but at least one users Emacs identifies itself as just "N.M",
so stop assuming that the version string always consist of at
least three elements.
Fixes #33.
|
|
|
|
|
|
| |
This makes it easier to make changes without having to recompile
all packages that use the `with-editor' macro, and the result of
macro-expansion gets smaller too.
|
| |
|
|
|
|
| |
...instead of `error' or `message'.
|
| |
|
| |
|
|
|
|
| |
Closes #32.
|
|
|
|
|
|
|
|
| |
If `*-1' fails, then we have to return nil. To make it clear
that this is what happens, show the return value (nil) before
the sexp-wise confusing multi-line warning text.
Also remove the unnecessary `format'.
|
| |
|