org-capture, when used on a target that doesn't exist, will invisibly
expand a file template, leaving a hang yasnippet active. After
org-capture initializes, the yasnippet will attempt to operate on
overlays that don't exist, throwing overlayp errors.
CDing into the project root can be accomplished with external tools, e.g.
cd $(git rev-parse --show-toplevel)
cd $(hg root)
cd $(npm root)
Any of which could be aliased. Also, +vterm/toggle and term/toggle
define the PROOT environment variable, so `cd $PROOT` will work too.
On the other hand, CDing to the current file/folder requires that the
shell be made aware of the file/directory of some Emacs state, which is
a little trickier to deal with, so I made that the default behavior for
+term/toggle, +vterm/toggle and +eshell/toggle.
Uses the most basic, uncustomized shell to a) prevent interference
caused by slow shell configs and b) speed up project text searches. This
comes at the cost of isolating these programs from envvars that the user
may have set in their shell configs in order to change
ag/rg/pt/git-grep/grep's behavior.
If this bothers you, change +ivy-file-search-shell to your shell (or
to the value of `shell-file-name`).
Possibly relevant to an issue mentioned in #1260
They are:
]m, [m
Jump to next/previous beginning of method/function.
]M, [M
Jump to next/previous end of method/function
]#, [#
Jump to next/previous preprocessor directive (only supports C-style
directives for now)
]*, [* (or ]\, [\)
Jump to next/previous comment
- Move subr-x/cl-lib loading to core-lib
- Revise docstrings for and rename various CLI functions to be more
descriptive and up-to-date
- After regenerating autoloads file, bin/doom will try to reload
autoloads files remotely, through the server/daemon, if possible. This
is highly experimental and could break
The fix for #1489 (in 13f5a762) gives the temp buffer (where formatting
is performed) a real buffer-file-name, which causes it to prompt to save
it when it is killed.
Adds these keys, inspired by tpope/vim-unimpaired:
[ SPC, ] SPC (or [o, ]o)
Add COUNT blank lines above/below the cursor
[f, ]f
Visit previous/next file in current folder, alphabetically
[u, ]u
Url encode/decode operators
[y, ]y
C-string-style escaping/unescaping operators (escapes quotes,
backslashes and control characters)
[x, ]x
XML entity encoding/decoding operators (only if :lang web is enabled)
[F, ]F
Focus previous/next frame (decided this is better than ]t/[t which is
being used by hl-todo-{next,previous})
We already had gp and ]b/[b (buffers).
We're not going to port ]e/[e because it is redundant with ddp/ddP or gx
(evil-exchange). I also think these keybinds are better suited to
{next,previous}-error.
multi-term performs a lot of hackery on Emacs' windows API, which causes
problems, particularly for persp-mode. To side-step it all we must
reinvent some of the wheel.
Fixes:
- Wrong-type-argument error when failing to display eshell buffer with
pop-to-buffer, because it was killed (when prefix arg is non-nil).
- Passes non-existent variable `buf` to +eshell-run-command.
Sets out to solve a number of issues with the package management
process. Namely:
- To-be-removed packages that are simply being removed are no longer
incorrectly labeled "quelpa->elpa", but "removed" instead.
- A backend (elpa vs quelpa) column was added to the package listing
confirmation when running `doom update`.
- Doom now correctly recognizes that packages installed with a psuedonym
are installed, and will not endlessly attempt to uninstall and
reinstall them on every `doom refresh`.
- Packages declared with :built-in will no longer lose their built-in
marking if said package is not actually present in Emacs' site load
paths. i.e. if you say it's built in, Doom won't question it.
- package!'s :ignore property is now treated as a form whose evaluated
result will be used as its value.
