Re: [RFC PATCH v1 3/6] kernel-doc: add kerneldoc-lint command

From: Jani Nikula
Date: Wed Jan 25 2017 - 03:22:11 EST

On Wed, 25 Jan 2017, Daniel Vetter <daniel@xxxxxxxx> wrote:
> On Tue, Jan 24, 2017 at 08:52:41PM +0100, Markus Heiser wrote:
>> this patch adds a command to lint kernel-doc comments.::
>> scripts/kerneldoc-lint --help
>> The lint check include (only) kernel-doc rules described at [1]. It
>> does not check against reST (sphinx-doc) markup used in the kernel-doc
>> comments. Since reST markups could include depencies to the build-
>> context (e.g. open/closed refs) only a sphinx-doc build can check the
>> reST markup in the context of the document it builds.
>> With 'kerneldoc-lint' command you can check a single file or a whole
>> folder, e.g:
>> scripts/kerneldoc-lint include/drm
>> ...
>> scripts/kerneldoc-lint include/media/media-device.h
>> The lint-implementation is a part of the parser module (
>> The comandline implementation consist only of a argument parser ('opts')
>> which calls the kernel-doc parser with a 'NullTranslator'.::
>> parser = kerneldoc.Parser(opts, kerneldoc.NullTranslator())
>> Latter is also a small example of how-to implement kernel-doc
>> applications with the kernel-doc parser architecture.
>> [1]
>> Signed-off-by: Markus Heiser <markus.heiser@xxxxxxxxxxx>
> Didn't we have a patch from Jani to gives us a make target doing this? I
> think that'd be even neater ...

Yes, see below. It's simplistic and it has an external dependency, but
it got the job done. And it does not depend on Sphinx; it's just a
kernel-doc and rst lint, not Sphinx lint. Whether that's a good or a bad
thing is debatable.

Anyway, I do think the approach of making 'make CHECK=the-tool C=1' work
is what we should aim at. Markus' patch could probably be made to do
that by accepting the same arguments that are passed to compilers.