Re: [PATCH v2 1/2] scripts: add tool to run containerized builds

[Nathan Chancellor <nathan@kernel.org>]

On Sun, Dec 21, 2025 at 09:09:17PM +0100, Guillaume Tucker wrote:
> On 19/12/2025 10:27 pm, Nathan Chancellor wrote:
> > On Thu, Dec 18, 2025 at 01:49:52PM +0100, Guillaume Tucker wrote:
> > ...
> > > +    def __init__(self, args, logger):
> > Adding something like
> > 
> >      self._args = [
> >          '--rm',
> >          '--tty',
> >          '--volume', f'{os.getcwd()}:/src',
> >          '--workdir', '/src',
> >      ]
> > 
> > here then adding an __init__() in the subclasses to append the runtime
> > specific arguments would allow _do_run() to be moved into
> > ContainerRuntime(). Otherwise, this looks pretty good and extensible.
> 
> Yes, I left these very similar parts as-is on purpose to make it very
> clear what the command line arguments are for each container runtime.
> It's a good idea to refactor this though as you mention, and I'll add
> a debug log message to print the command line instead.
> 
> I'm also anticipating that other runtimes will be quite different,
> Podman and Docker just so happen to have many options in common.
> Things like runc or containerd are very different beasts so I don't
> want to over-generalise.  But this is straightforward enough for now.

Yeah, I had figured that might have been the reason for keeping these
things separate, which does make sense. At the same time, I would rather
keep things simple and shared now since we can with no other runtimes
supported. If we want to add new runtimes in the future, undoing this
should be simple enough and make it obvious to see why the separation is
happening.

> So I'll rework this a bit in a v3.

Thanks a lot!

Cheers,
Nathan