On Thu, Feb 07, 2019 at 06:33:34PM +0100, Markus Heiser wrote: > > Am 07.02.19 um 17:18 schrieb Mike Rapoport: > >>>Does checkpatch checks the kernel-doc parts at all? > >>No. I guess there are to many places to fail / to hard to put someone in > >>charge. E.g. if you do include a single kernel-doc comment from a source all > >>kernel-docs in the source will be parsed and may produce (error/warning) > >>essages. What we have, are some targets: > >> > >>-linkcheckdocs > >> check for broken external links (will connect to external hosts) > >> > >>- refcheckdocs > >> check for references to non-existing files under Documentation > >Right, but these should be checked explicitly and I doubt many people do it > >before submitting patches. OTOH, checkpatch is something that's widely used > >and if it had verified the kernel-doc parts, more comments would be > >following the convention. > > I'am with you, but I do not have any clue how to solve this Gordian Knot > faithful and without massive collateral damage / sorry :| > > The only thing I know, we have the -none option: > > $ ./scripts/kernel-doc -none ./include/media/cec.h > ./include/media/cec.h:51: warning: Function parameter or member 'lock' not > described in 'cec_devnode' > > But this is nothing more than noise if the patch does not touch cec_devnode. > And there is another problem I see, if we want to check refs ... Well, the case when a patch changes function parameters but forgets to update the kernel-doc part is particularly annoying. I believe it's possible to match function parameter changes with the corresponding kernel-doc changes (or lack of them). > >> -linkcheckdocs > >> check for broken external links (will connect to external hosts) > >> > >> - refcheckdocs > >> check for references to non-existing files under Documentation > > The refs are solved late in the sphinx build process when .rst files and > kernel-doc comments come together .. so we need sphinx for checkpatch, > I gues this is a no-go (?) > > -- Markus -- > -- Sincerely yours, Mike.
