Re: [PATCH linux-kselftest/test v6] lib/list-test: add a test for the 'list' doubly linked list

[Brendan Higgins <brendanhiggins@google.com>]

On Wed, Oct 30, 2019 at 10:18:44AM -0700, Joe Perches wrote:
> On Wed, 2019-10-30 at 09:35 -0700, Brendan Higgins wrote:
> > Agreed. I can see the point of not wanting to write an exception into
> > checkpatch for every exception of it's general rules; however, it
> > would be nice if there was a way to maybe have a special comment or
> > something that could turn off a checkpatch error. That way, a
> > checkpatch error/warning always means some action should be taken, and
> > if a rule is being ignored, there is always documentation as to why.
> 
> That couldn't work when a comment which may exist
> in a file is out of scope of the patch context.

Sorry, I don't understand exactly what you mean. Can you elaborate?

If it wasn't obvious, I am not proposing that David should make the
changed I described now for this patch. I know what I proposed would not
be an easy thing to implement, especially given the opinions that it is
likely to solicit.

Nevertheless, in the long term, I have seen other projects allow a
comment that would cause style checkers or static analysis tools to
ignore the designated line. Maybe we could implement this as a line
comment that suppresses a checkpatch warning of a certain kind on the
line. So here, we might have something like:

static void list_test_list_for_each_prev(struct kunit *test) /* checkpatch: disable=for-each-format */

We would also probably want to require an explanation either in the
checkpatch comment or the line above, but then you have to worry about
that comment not being included in a patch that only updates the
offending line.

Anyway, it's just an idea. I know that we don't currently assume that
all checkpatch errors/warnings require some action, but it might be cool
if they did.

Cheers