| Age | Commit message (Collapse) | Author |
|---|
|
Move Python settings to use-package. Tweak a couple of knobs while in
there.
|
|
|
|
|
|
|
|
* "Extra keys" are advertised for non-graphical configurations; as it
happens, they are also useful to provide substitutes for bindings
based on arrow keys.
* Speed commands are nice. Might tweak the predicate some day, but
the default is already useful.
|
|
|
|
Also rename theme commands.
|
|
* "Regular" backgrounds need to pop out more against "dim" ones: this
makes refinements easier to spot within "focused hunks" in Magit.
* Also make dim & subtle foregrounds more saturated. That gives
"poorer" contrast, but it's still legible, and "poorer" contrast is
sort of a design goal for the "dim" variant.
|
|
|
|
… I should go back to Modus.
|
|
* magenta/red is "the current thing"
* cyan is "other things elsewhere that match the thing"
|
|
For stuff that's interesting, but not the main event.
|
|
This came in handy in the revamp soon to be committed.
|
|
This declines a hue for both background and foreground use.
|
|
Mostly to avoid icomplete jank (bug#40152).
Inspired by:
<https://www.scss.tcd.ie/~sulimanm/posts/default-emacs-completion.html>
Rationale for the customizations:
C-M- chord for navigation Because I found myself missing C-n/C-p in
a couple of situations.
auto-help 'visible The feedback is not as immediate as
icomplete, but it's good enough. Tried
'always, but there's a sit-for somewhere
that causes a weird pause if you happen to
RET before bringing up completions.
auto-select 'second-tab M-v? M-g M-c??
show-help nil "Click on a selection to select it" 😏
format 'one-column Where have you been all my life.
max-height 10 More often than not I'm typing something
inspired by the content from another
buffer, so limit how much
context *Completions* can hide.
auto-choose icomplete never clobbered the minibuffer
until I asked it to (with e.g. C-M-i).
Since C-u M-RET is a thing, keep
candidates off the minibuffer unless I
pull them in, in case I change my mind and
start typing something entirely different
from the currently highlighted candidate.
|
|
Puzzled by this problem I have with :hook. "(use-package) Hooks"
suggests that if package X defines function X-foo to be used by
package Y in Y-bar-hook, then the idiomatic thing to do is…
(use-package X
:hook ((Y-bar . X-foo)))
… but empirically, if Y.el contains…
(defcustom Y-bar-hook '(Y-quux Y-corge))
(add-hook 'Y-bar-hook #'Y-grault)
… then Y-bar-hook will be set to '(X-foo Y-grault): it will be missing
the functions added in the defcustom form.
Since add-hook does this:
(or (boundp hook) (set hook nil))
I am assuming that the problem happens when use-package runs X's :hook
additions before Y is loaded. No idea (a) if I can reproduce from -Q
(b) if this is a SNAFU from setting always-defer (c) if this could be
solved by sprinkling :after or :requires.
|
|
|
|
Make region cyan rather than blue, so that the selection remains
visible over added hunks from diff or Magit buffers.
Make Magit selections cyan to follow suit. Make magit-blame headers
blue to fit the "intangible UI" motif.
Look longingly at Modus 4.2 announcement and wonder if this is all
worth it.
|
|
Recent ElDoc & Eglot developments have made the defaults much more
tolerable (the echo area height does not oscillate as much) &
useful (linter warnings are no longer preempted by symbol
documentation).
|
|
|
|
'cause why not.
|
|
|
|
|
|
* ponder magical "kill context so I can paste it into another app"
commands,
* remove obsolete Magit variables,
* make org-goto use the "path completion" UI,
* ponder better buffer names for some buffers.
|
|
|
|
|
|
|
|
|
|
Ideally I'd like Emacs to stop highlighting subcommands altogether,
since it can't properly figure out nested commands, and it does not
append sh-quoted-exec to whatever face is already there, e.g. when
writing a subcommand in a string "$(like this)".
|
|
|
|
|
|
Vector 2022 👏
|
|
* Make header names more "prompt-like" rather than "type-like".
* Tone down header content.
* Remove color for non-MIME attachments.
|
|
|
|
It's going to keep bothering me otherwise.
|
|
OT1H it's likely I'll always keep font-lock's function & built-in
faces somewhat close together (although thinking more about it, the
former is for highlighting declaration, and the latter for
highlighting usage, so… worth challenging that assumption), OTOH I'd
rather not keep the implicit function = directory association.
|
|
Leave font-lock faces alone; just unset shell-font-lock-keywords.
|
|
The first mistake here was to forget to wrap console codes with \[…\];
for some reason though after adding those…
export PS4=$(echo -en '\[\e[7m\]+ \[\e[0m\]')
… bash no longer repeats the first character (+) according to the
indirection level, or rather, it repeats an entirely unrelated
character:
$ ( PS4='+ ' ; set -x ; echo $(echo $(echo hi | grep hi)) )
+++ echo hi
+++ grep --color=auto hi
++ echo hi
+ echo hi
hi
$ ( PS4=$(echo -en '\[\e[7m\]+ \[\e[0m\]')
set -x
echo $(echo $(echo hi | grep hi))
) |& hexdump -C
00000000 01 01 01 1b 5b 37 6d 02 2b 20 01 1b 5b 30 6d 02 |....[7m.+ ..[0m.|
^^ ^^ ^^ (echo)
00000010 65 63 68 6f 20 68 69 0a 01 01 01 1b 5b 37 6d 02 |echo hi.....[7m.|
^^ ^^ ^^ (grep)
00000020 2b 20 01 1b 5b 30 6d 02 67 72 65 70 20 2d 2d 63 |+ ..[0m.grep --c|
00000030 6f 6c 6f 72 3d 61 75 74 6f 20 68 69 0a 01 01 1b |olor=auto hi....|
^^ ^^ (echo)
00000040 5b 37 6d 02 2b 20 01 1b 5b 30 6d 02 65 63 68 6f |[7m.+ ..[0m.echo|
00000050 20 68 69 0a 01 1b 5b 37 6d 02 2b 20 01 1b 5b 30 | hi...[7m.+ ..[0|
^^ (echo)
00000060 6d 02 65 63 68 6f 20 68 69 0a 68 69 0a |m.echo hi.hi.|
0000006d
No idea where those ^A characters come from.
While at it, wrap PS2 to avoid line-wrapping shenanigans.
|
|
* inputrc goodies,
* let readline see C-s,
* reverse PS4 for kicks,
* protect crippled shells (M-x shell, batch ssh commands) from this
savagery.
|
|
|
|
|
|
|
|
|
|
Highlights:
* add etc/NEWS tags
* some Markdown markup
* tone done fringe bitmaps
* make current icomplete selection easier to spot on monitors with
poor contrast for black levels
|
|
|
|
|
|
* Make region more prominent. The main blocker for that was
lazy-highlight's cyan hues being too close, but now that it has
:inverse-video, s'all good.
* Make subgroup matches more prominent. Go with red & blue to "frame"
magenta.
* Add dim backgrounds to (Org, Markdown) code snippets, to mimic that
"inset" effect other applications sometimes apply to those.
* Remove green foreground to ~Org inline code~; let background and
serifs speak for themselves.
* Crank up saturation on =Org verbatim= `Markdown inline code` to
compensate lighter background.
|
|
|
|
show-paren-mode is not needed for Emacs≥28, but I still have stuff in
these dotfiles that purport to support older versions.
|
|
Wonder how long I can live with that "make commit summaries all
title-like" gag.
|
