Re: [PATCH] checkpatch: warn for p0 patch only if prefix is not b

[Dafna Hirschfeld <dafna.hirschfeld@collabora.com>]

On 11.01.22 13:27, Lukas Bulwahn wrote:
> On Tue, Jan 11, 2022 at 11:47 AM Dafna Hirschfeld
> <dafna.hirschfeld@collabora.com> wrote:
>>
>>
>>
>> On 11.01.22 12:18, Lukas Bulwahn wrote:
>>> Dafna,
>>>
>>> On Mon, Jan 10, 2022 at 9:06 PM Dafna Hirschfeld
>>> <dafna.hirschfeld@collabora.com> wrote:
>>>>
>>>> It might be that file 'b' happens to exit. In that
>>>> case, if the prefix is also 'b' (which is the
>>>> common case) we get the falsely warning:
>>>>
>>>> patch prefix 'b' exists, appears to be a -p0 patch
>>>>
>>>> So warn only if prefix is not 'b'
>>>>
>>>
>>> The checkpatch script that is maintained here is really only intended
>>> for its use in the kernel development. You may use checkpatch anywhere
>>> else, but any changes that increase complexity for those other use
>>> cases is really difficult to argue for inclusion in the kernel
>>> repository. The checkpatch script currently is already large and
>>> complex enough and all rules need to be understood as rough
>>> heuristics, not as strict rules.
>>>
>>> So, can you point to a kernel repository where there is actually a
>>> file 'b' included? On a quick scan, I could not find a file 'b' in the
>>> current trees of the repositories on my machine.
>>>
>>> I am just letting you know about what I have observed; I do not decide
>>> on the inclusion of this patch, though.
>>
>> Hi, a 'b' file might make it to the source folder as an untracked file.
>> This actually happened to me since I was too lazy to give it a meaningful name.
>> Then I got this warning and it took me some time to figure out what is the problem.
>>
> 
> Well, but you run checkpatch.pl on a patch, right? So, you need to add
> the file explicitly with git (where you notice adding a file called b,
> which probably is really not a good name), you create a git commit
> (where that is pointed out again), then create a patch from that
> (which you may manually look at again) and then run checkpatch.pl
> before you submit it (again, submitting a patch with a file 'b' is
> probably a good reason to rethink your submission).

Hi , no, the 'b' file is left untracked.
It is easily reproducible, inside a kernel source repo do:

touch b

# These two commands will both trigger the warning.
./scripts/checkpatch.pl --strict -g HEAD
./scripts/checkpatch.pl --strict 0001-some-random-patch.patch

> 
> If it helps, you can add some documentation on the PATCH_PREFIX rule
> in the checkpatch documentation at
> ./Documentation/dev-tools/checkpatch.rst. Especially, you can note the
> situation you encountered there, e.g., that adding files with explicit
> name 'a' or 'b' may make this rule trigger. If that documentation of
> the rule is helpful, I will ack that documentation patch and request
> inclusion of it.

I can do that as well if you think it is better.

Thanks,
Dafna

> 
> Lukas
> 
>> Thanks,
>> Dafna
>>
>>>
>>> Lukas
>>>
>>>> Signed-off-by: Dafna Hirschfeld <dafna.hirschfeld@collabora.com>
>>>> ---
>>>>    scripts/checkpatch.pl | 2 +-
>>>>    1 file changed, 1 insertion(+), 1 deletion(-)
>>>>
>>>> diff --git a/scripts/checkpatch.pl b/scripts/checkpatch.pl
>>>> index 1784921c645d..72263b142e39 100755
>>>> --- a/scripts/checkpatch.pl
>>>> +++ b/scripts/checkpatch.pl
>>>> @@ -2821,7 +2821,7 @@ sub process {
>>>>                           $in_commit_log = 0;
>>>>
>>>>                           $p1_prefix = $1;
>>>> -                       if (!$file && $tree && $p1_prefix ne '' &&
>>>> +                       if (!$file && $tree && $p1_prefix ne '' && $p1_prefix ne 'b' &&
>>>>                               -e "$root/$p1_prefix") {
>>>>                                   WARN("PATCH_PREFIX",
>>>>                                        "patch prefix '$p1_prefix' exists, appears to be a -p0 patch\n");
>>>
>>>> --
>>>> 2.17.1
>>>>