On 2021/09/03 16:31, Frederic Cambus wrote:
> On Mon, Aug 12, 2019 at 07:39:23AM +0200, Theo Buehler wrote:
> > On Sun, Aug 11, 2019 at 11:30:09PM +0100, Stuart Henderson wrote:
> > > So the claimed advantage is that it's super fast, but actually it uses way
> > > more cpu than probably any other terminal emulator in the tree except
> > > cool-retro-term?
> >
> > Yes, it is blazing fast. matthieu's test of 'cat /etc/termcap' completes
> > in under 0.1s while xterm uses 0.2-0.3s. It does feel way snappier than
> > xterm in general.
> >
> > It looks like I made CPU use worse by about a factor of 2 due to using
> > vm.malloc_conf=CFJ. Still, ~5% is way more than it should need.
> >
> > Another advantage is that contrary to xterm it is trivial to configure
> > thanks to a self-documenting .yml config file.
> >
> > > Is this useful enough to be worth adding so much time to a bulk build?
> >
> > I doubt it.
> >
> > > It will likely need to be amd64-only.
> >
> > Probably. Testing on other archs would need more time than I currently
> > have.
>
> Resurrecting this as it seems the requirements are much more reasonable
> nowadays, and I can build the latest version in 4 minutes on my Ryzen 5
> PRO 2500U laptop:
>
> Finished release [optimized] target(s) in 4m 07s
>
> Here is a new tarball with the following changes:
>
> - Update to alacritty 0.9.0
> - Move MODCARGO_CRATES directives into crates.inc.
> - Remove the ulimit checks, as I was able to build the port as a regular
> user with the default limits without running out of memory
README needs either rewriting or removing. Some of the sentences don't
parse. The configuration section says what the program does but doesn't
really say what the user should do with that information.
There's a conflict between what DESCR says ("It provides sane choices
for defaults and requires no additional setup") and README: "Although
the default configuration work on OpenBSD, some values are not optimal
[...] This option leads alacritty to a non negligeable CPU overhead on
OpenBSD". If it's bad enough to document then I would suggest patching
the default setting for live_config_reload instead. And rather than
telling the user "please install pkg_add xcursors-themes port" it would
be better to just RUN_DEPENDS on it. If that's done then there isn't
really any more need for the pkg-readme, the standard documentation
should suffice.
