Setting a face's `:font` attribute implicitly sets a host of other faces
attributes (:family, :foundry, :width, :weight, :height, and :slant),
which is problematic in places where these faces are used in tandem with
other faces, like how EWW renders bold elements with both
variable-pitch+bold faces, with the expectation that their attributes
would cascade properly, but not so if variable-pitch sets :weight or
:slant.
We can't be certain where MELPA/ELPA packages come from, but
occasionally, their hosted on unstable servers (savannah) or git servers
with certain features disabled (as is the case for paredit's not
allowing shallow clones).
Trust in emacsmirror, and we can be more certain the package is coming
from github.
This may trigger a rebuild of your packages!
Reverts 86fd6c621 which increased gcmh-high-cons-threshold from Doom's
16mb default to its upstream default (of 1gb). This was an experiment.
In theory, the user should idle enough that it would GC enough to stave
off freezes/stuttering/paging, but in practice, either users did not
idle enough or the GC didn't clean up enough when given the opportunity.
The result: terrible stuttering and freezing after long periods of use.
Back to 16mb gang.
There are legitimate reasons why a user would want to treat $HOME as a
project. 'doom doctor' now complains about this case in greater detail.
I'll leave it to users to deal with this edge case.
Reduces the amount of "noise" included in bin/doom's output.
Also fixes an issue where warnings during autoloads generation would
sneak into Doom's autoloads file, producing weird void-variable errors,
like
(void-variable . rainbow-delimiters:)
(void-variable . diredfl:)
(void-variable . company:)
Its output is helpful. Let's not silence it.
But prevent output during incremental loading from hijacking the
minibuffer (likely the original reason org-roam-verbose was disabled).
Causes a regression where output was written into /tmp/doom.sh,
producing errors like:
/tmp//doom.sh: string 4: $'\E[32m✓': command not found
I'll revert this until I have a better solution.
Half-reverts 628f0a9, which forced straight to byte-compile packages in
another Emacs session, which produced failures for packages that didn't
properly load their compile-time dependencies (e.g. macro calls).
But now, I realize that _not_ failing in those cases is worse and
produces odd byte-code issues like #1657 or invalid-function ((date
date)) errors. This doesn't guarantee a fix, but at least takes the ball
out of Doom's court, and gracefully fails to uncompiled packages, rather
than incorrect bytecode.
We only need this magic in CLI sessions. It's better to only use half
the CPUs in interactive sessions (if the user has enabled
comp-deferred-compilation for some reason).
Fixes#5042
With a commit of e2a11d24fd, when it comes to
`doom compile`, modules aren't loaded anymore, which corrupts the compiling process.
What I did is just adding the necessity parts for the doom's byte compiler.
