unicode-fonts takes 5-15 seconds to remap unicode blocks on first
invokation, but it does so invisibly because inhibit-redisplay and
inhibit-message are active during startup (to prevent unintended redraws
slowing down the startup process). This change ensures users get
meaningful feedback during this time.
BREAKING CHANGE: alchemist appears to be unmaintained and has reportedly
been replaced by elixir-ls and LSP (#5488). To restore your Elixir dev
environemnt, enable :lang (elixir +lsp) and install elixir-ls via your
package manager or 'M-x lsp-install-server'.
Close: #5488
- Renames "Doom Emacs issues" to "Doom issues" for brevity in the
completion window.
- Adds discourse.doomemacs.org search provider.
- Adds Internet Archive search provider.
This would produce extra (and confusing) noise in envrc's error popup,
and also means multiple (potentially expensive) calls to direnv in
failure cases.
Compacting elfeed's DB when killing the search buffer can incur a long
delay (which blocks Emacs). Doing so isn't really critical, so I've
moved it to kill-emacs-hook instead.
Close: #7127
Co-authored-by: merrickluo <merrickluo@users.noreply.github.com>
Org complains if org-loaddefs.el is missing, but Straight generates a
org-autoloads.el instead (and loads it separately), so we need only fool
Org it exists.
This breaks Doom in scenarios where the user has set a non-standard
doom-localleader-key, because the module is binding a localleader key
too early in the startup process.
Amend: ca90c5e83cFix: #7130
Ref: #7136
This is to bring the `sh-shell` variable in scope, when it otherwise
would not be if a `Sh` buffer hadn't yet been opened in the current
Emacs session.
After some experimentation, calling `sly` non-interactively when no
Slynk connection involves some strange async affects which displease the
REPL-opening code. Instead, we start the Slynk process, wait while it
loads, and then finally connect with the usual `sly-mrepl`, whose Elisp
would have been loaded by this point.
Previously, when setting up a prompt for choosing from the available
REPLs, not all known options were detected via the regex pattern. This
commit modularizes the logic and provides a backup for detecting REPLs
that we definitely know about from the `+eval-repls` list.
These two can *significantly* slow down larger org buffers for evil
users, when switching modes (e.g. leaving insert/replace mode), so I am
removing these if/when I find a better alternative. Though, they can
still be done manually with `C-c C-c` (for cookies) and `TAB` in tables.
From now on, our documentation will assume your Emacs config lives in
~/.config/emacs, by default, rather than ~/.emacs.d. Support for the
latter is not going away, it will simply be mentioned less in the
literature, as all supported versions of Emacs going forward (and future
versions of Doom) will support (and prefer) XDG conventions.
The user manual will be updated separately.
Close: #6965
Co-authored-by: gagbo <gagbo@users.noreply.github.com>
A warning gets logged if a lazily-loaded evil-collection package is
loaded too early at startup, but the goal posts were in the wrong place,
resulting in unavoidable warnings for packages that weren't actually
loaded too early.
In the Elder Days, it was common to give Fortran files the extensions of
`.FOR` or later `.F90`. The major modes for Fortran don't automatically
detect this conversion. The former was already accounted for in Doom's
module, but not the latter. This commit rectifies this.
