Re: [RFC PATCH] checkpatch: correctly detect lines of help text

[Randy Dunlap]

On 12/2/20 10:54 AM, Joe Perches wrote:
> On Wed, 2020-12-02 at 19:27 +0100, Nicolai Fischer wrote:
>> Currently, checkpatch uses keywords to determine the end
>> of a Kconfig help message which leads to false positives:
>>

> 
> I believe all the '---help---' lines have been converted to just 'help'
> so the '(?:---)?' bits here could be removed.

Yes.

> See:
> 
> commit 22a4ac026c15eba961883ed8466cb341e0447de1
> Author: Masahiro Yamada <masahiroy@xxxxxxxxxx>
> Date:   Wed Jun 17 12:02:20 2020 +0900
> 
>     Revert "checkpatch: kconfig: prefer 'help' over '---help---'"
>     
>     This reverts commit 84af7a6194e493fae312a2b7fa5a3b51f76d9282.
>     
>     The conversion is done.
>     
>     Cc: Ulf Magnusson <ulfalizer@xxxxxxxxx>
>     Signed-off-by: Masahiro Yamada <masahiroy@xxxxxxxxxx>
> 
[]
> 
> Indentation can vary in the help blocks.  For instance:
> 
> arch/Kconfig:   help
> arch/Kconfig-     Functions will have the stack-protector canary logic added in>
> arch/Kconfig-     of the following conditions:
> arch/Kconfig-
> arch/Kconfig-     - local variable's address used as part of the right hand sid>
> arch/Kconfig-       assignment or function argument
> arch/Kconfig-     - local variable is an array (or union containing an array),
> arch/Kconfig-       regardless of array type or length
> arch/Kconfig-     - uses register local variables
> arch/Kconfig-
> 
> This doesn't allow blank lines for multi-paragraph help text either.
> 
> I think keyword parsing is necessary and some false positives are
> inevitable as the parsing logic in a line-by-line analyzer will
> always be incomplete.


There are also large hunks of block/Kconfig and drivers/hid/Kconfig
that don't use any indentation for help text...
in case that matters here.


-- 
~Randy