[Linux-kernel-mentees] git_commit_id false positives
dwaipayanray1 at gmail.com
Tue Nov 10 20:18:33 UTC 2020
>> First of all there are 45 warnings of type:
>> References: 2248a28384fe ("drm/i915/gen8+: Add RC6 CTX corruption WA")
>> References: 860afa086841 ("drm/i915/gt: Flush gen7 even harder"), etc.
>> Should references be simply exempt from this check? I don't
>> know much about it. But seems checkpatch is incorrect here.
> I understand what the authors intend to do. They want to refer to a commit as a tag, such as Link: for URLs or Fixes: for commits that fixed another commit. However, this format of using References: is not an agreed and common convention.
> We could ask on the mailing list if there is some preference for such a common agreement. Then, we can extend the kernel process documentation and add a rule for checking this convention in checkpatch.
> Do you want to do that? Expect some heavy discussion and some bikeshedding.
That certainly sounds intimidating to me. But I would like to give it
a go if possible. Whom do I mail for that? The documentation mailing
list or somewhere else?
>> And finally the other fixable warning is for those with split commit
>> titles itself.
> Yes, you saw that another mentee started this work but could not really finish that (yet?), right?
> This sounds like a good case.
Sure then I will work on a fix for this.
>> Other than that, there was also this particular case with non escaped
>> quotes within the commit title.
>> commit 97a32539b956 ("proc: convert everything to "struct proc_ops"")
>> Should this be handled as well. Or is this a human error? I am not quite
> Of course, a commit title can contain quotes itself, but checkpatch does not handle that case properly; this case can be quite tricky to handle. Let us fix the other cases first and then see with an evaluation how often this remaining case really happens.
Okay I will do that.
> Another case of another non-standard commit reference is that when I refer to a commit with a broken title that still contains "[PATCH] xyz: do something" in the git history, I simply refer to it only with commit abcdabcdabcd ("xyz: do something") and I drop the [PATCH] prefix.
> That is a non-standard way but I think it should be accepted as standard or at least acceptable to checkpatch. Again, that also needs some discussion with more kernel developers.
I didn't know about this one. But I will try to get an
estimate on it and probably a fix.
More information about the Linux-kernel-mentees