Fixes an issue where autoloads would bloat load-path and auto-mode-alist
with duplicates because they were build on top of old autoloads state.
We initialize packages manually at the start of non-interactive sessions
anyway.
- No need to use after! (it has a little magic associated with it that are
meaningless this early in the startup process), so we use
with-eval-after-load instead.
(And I've been itching for an excuse to use doom-rpartial)
This update addresses two evasive issues:
1. Packages updated with `doom update` would not rebuild correctly,
requiring a `doom refresh` afterwards,
2. Packages would fail to rebuild even if their byte-compiled files were
stale. The result: "*.el is newer than *.elc" warnings at startup.
This update may potentially break your usage of add-hook! if you pass
the :local or :append properties to it. This is how they used to work:
(add-hook! :append 'some-mode-hook #'do-something)
Thsoe properties must now follow the hooks, e.g.
(add-hook! 'some-mode-hook :append #'do-something)
Other changes:
- Various add-hook calls have been renamed to add-hook! because I
incorrectly assumed `defun` always returned its definition's symbol,
when in fact, its return value is "undefined" (so sayeth the
documentation). This should fix#1597.
- This update adds the ability to add multiple functions to hooks
without a list:
(add-hook! 'some-mode-hook
#'do-something
#'do-something-else)
- The indentation logic has been changed so that consecutive function
symbols at indented at the same level as the first argument, but forms
are indent like a defun.
(add-hook! 'some-mode-hook
#'do-something
#'do-something-else)
(add-hook! 'some-mode-hook
(message "Hello"))
And ensure it runs as late as possible in after-change-major-mode-hook.
Hopefully to increase the probability of
doom-highlight-non-default-indentation-h detecting a user-enabled
whitespace-mode and bowing out in time.
`shell-command` erases the target buffer before piping output to it.
Which means the envvar file comment header was wiped out. This causes
trouble for the envvar file parser, which expects the envvar list to
start with two newlines.
Caused lecause straight is prompting for input in a headless
session *and* from a headless async child instance of Emacs. It waits
forever for input for a question we could never see or respond to.
How sad.
This will run the unit tests for each module in a separate Emacs
instance. It's a fair bit slower, but much more useful for something as
stateful as an Emacs config.
Now I just need to push the rewritten tests.
- Let-bind byte-compile-* vars instead of using file-local vars.
- Fix duplicate bullet point in "Copied backup..." message.
- Only display refresh message if cli command was successful.
Straight expects to be used interactively, which don't do (yet). Its
transactional system depends on idle timers, which don't run in a
noninteractive session, so we have to nudge it ourselves.
- Now handles errors from threads gracefully, rather than failing
silently.
- Exploits straights modification system to trigger rebuilds
later (instead of force-rebuilding after each update).
