use-package specified lisp-indent-function to indent like defun.
Currently, use-package main macro specify indent as (declare (indent 1)),
then change indent mode like defun.
(declare (indent defun)) is same effect.
And it is useful when redefining use-package.
Emacs 25 defined a global variable `features', which triggers a
warning "Lexical argument shadows the dynamic variable features".
That's not `use-package's fault, but we should suppress the warning
anyway, so that there is no additional noise that would cause us to
potentially overlook warnings that absolutely have to be addressed.
:ensure-system-package was installing packages by running
system-packages-get-command via async-shell-command. This meant that
system-packages-use-sudo wasn't being honoured.
This patch makes :ensure-system-package use system-packages-install
for all cases, except where a custom install command is supplied, in
which case async-shell-command is used.
This issue was introduced in 9f034a0bcf [https://github.com/jwiegley/use-package/issues/673], as a fix for
[https://github.com/jwiegley/use-package/issues/661]. Prior to that commit, system-packages-use-sudo was being
honoured.
This patch also fixes a bug where a cons containing a lone symbol in a
list of conses causes nil to used as the package to install.
GitHub-reference: fix https://github.com/jwiegley/use-package/issues/720
Copyright-paperwork-exempt: yes
When it is at the end, the keys don't get bound until after the package has been loaded, which is bad. Placing it at the beginning allows the key bindings to be mapped to auto-loaded functions before the package is loaded, so that a key chord press triggers the package load.
By adding events of the form (cons t event) to unread-command-events, Emacs
correctly updates this-command-keys for the subsequent use of the prefix.
Before this change, packages like my which-key were not getting the correct
prefix after use-package-autoload-keymap loaded the relevant package.
See https://github.com/justbur/emacs-which-key/issues/192
The function introduced in https://github.com/jwiegley/use-package/issues/673 wasn't declared at compile time, and
it made byte compiler unhappy. Moreover, it was forgotten to remove
redundant compile time variables. Thanks @tarsius for pointing that
out.
This change
* Removes redundant variable declarations
* Adds `system-packages-get-command` function declaration.
Copyright-paperwork-exempt: yes
float-time expect list like (HIGH LOW USEC PSEC)
> HIGH has the most significant bits of the seconds, while LOW has the
> least significant 16 bits. USEC and PSEC are the microsecond and
> picosecond counts.
by `current-time` in editfns.c
ensure-system-package doesn't honor system-packages
customizations (https://github.com/jwiegley/use-package/issues/661), because system-packages didn't provide an API
to retrieve shell command to be executed.
This change makes use of the new system-packages'
`system-package-get-command` function and therefore fixes the issue.
GitHub-reference: fix https://github.com/jwiegley/use-package/issues/661
Copyright-paperwork-exempt: yes
bind-key supports the keyword ":package" but use-package does not know that. Adding this should be helpful. In the future, maybe we could automatically fill in with the use-package package, but that could be a breaking change.
Instead of testing the length of each form passed to :custom-face,
the sanity check would test the number of forms passed to :custom-face,
causing it to fail when more than 2 face customisations are used.
Fixes https://github.com/jwiegley/use-package/issues/600.
Copyright-paperwork-exempt: yes
The function `elisp--local-variables' inserts an unbound variable,
`elisp--witness--lisp', into macro forms to determine the locally bound
variables for `elisp-completion-at-point'. It ends up throwing a lot of errors
since it can occupy the position of a keyword (or look like a second argument to
a keyword that takes one). Deleting it when it's at the top level should be
harmless since there should be no locally bound variables to discover here
anyway.
I noticed that the "Gathering Statistics" instructions for rendering the Org
table could be automated for convenience. Now the table is reformatted
automatically, and you can still manually sort it. I'm not sorting it by default
to maintain the same order of use-package forms for easier troubleshooting.
