Re: Guix on the ASUS C201PA

[Timothy Sample <samplet@ngyro.com>]

Hi,

Ludovic Courtès <ludo@gnu.org> writes:

> Hello!
>
> Timothy Sample <samplet@ngyro.com> skribis:
>
>> I was able to get Guix to boot on an ASUS Chromebook C201PA.  This model
>> of computer is pretty neat.  It’s an ARMv7 (32-bit) machine that can be
>> run with entirely free software.  There is even a free graphics driver
>> in the works [1].
>>
>> I’ve attached a (messy) patch that adds a (hacky) bootloader definition
>> for Depthcharge and a Linux-Libre package that works on the machine
>> (using an unsupported version of Linux-Libre).  All those parenthetical
>> comments are supposed to suggest that this work is not really finished.
>> Now that the computer runs Guix, it should be straight-forward (if time
>> consuming) to fix some of these problems and arrive at something nice.
>
> Impressive work!

Thanks!

> The patch is really not what I expected given the qualifiers you gave
> above.  :-)  It could use more comments to explain what’s going on but
> apart from that it looks rather clean to my eyes.
>
> What do you think would be the best course of action to integrate it?
> Wait for Vagrant to test it and fix a couple of things?  ;-)  Or…?

On the bootloader side, if you are okay with the way it works, then it
is fine.  I would just add some comments.  Most importantly, I would
want to write somewhere that it is okay for the bootloader installer to
make use of the configuration file.  Beyond that, there are some
improvements that could be made, but they can come later.  For instance,
we could automatically roll-back one generation if booting fails.

For the kernel, the patch currently uses an unsupported version of
Linux-Libre, which is not great.  I checked that we can switch to the
4.19 version developed by the PrawnOS project.  However, they say that
Wi-Fi is broken on that branch (I don’t have a dongle yet, so I don’t
know).  Parabola has a working 5.0 kernel, but I’ve found that their
configuration is not as reliable as the PrawnOS one (it fails to boot
sometimes).  Ideally, we could figure out which configuration options
and patches are necessary and just augment what we already do, rather
than having a completely separate kernel package.  I’ve tried getting
there by trial and error, but that’s exhausting.  I’m hoping Vagrant’s
setup with the USB serial console will help, but I haven’t tried it yet.

In short, I would rather not rely so closely on what PrawnOS is doing.
It would be better to follow their approach loosely while using our own
kernel versions and configurations.  Does that make sense?


-- Tim