Re: [PATCH v2 1/2] kernel-doc: don't let V=1 change outcome

[Masahiro Yamada]

On Wed, Jun 7, 2023 at 6:07 AM Johannes Berg <johannes@xxxxxxxxxxxxxxxx> wrote:
>
> On Tue, 2023-06-06 at 20:15 +0900, Masahiro Yamada wrote:
> > >
> > >  ifneq ($(KBUILD_EXTRA_WARN),)
> > > -  cmd_checkdoc = $(srctree)/scripts/kernel-doc -none $<
> > > +  cmd_checkdoc = $(srctree)/scripts/kernel-doc -none \
> > > +        $(if $(KDOC_WALL), -Wall) \
> > > +        $(if $(KDOC_WRETURN), -Wreturn) \
> > > +        $(if $(KDOC_WSHORT_DESC), -Wshort-desc) \
> > > +        $(if $(KDOC_WSHORT_DESC), -Wcontents-before-sections) \
> >
> >
> >
> > Sorry, I misunderstood your intention.
> > (I just thought existing env variables would be moved to Makefile)
> >
> >
> > I do not want to proliferate env variables any more.
>
> Oh, ok, sure.
>
> > If you need per-flag control, maybe we can do like this?
>
> Well honestly, I myself just want to pass -Wall, but not necessarily W=2
> since that adds more stuff from the C compiler.
>
> > cmd_checkdoc = $(srctree)/scripts/kernel-doc -none \
> >               $(KDOCFLAGS)
> >
> >
> > Then, users can do
> >
> >   $ make KDOCFLAGS=-Wall
> >   $ make KDOCFLAGS=-Wreturn
>
> I'd rather call it KDOC_FLAGS if you don't mind to align with
> KDOC_WERROR which we have already, but sure, can do.


I just tried to be consistent with
CPPFLAGS, CFLAGS, AFLAGS, CHECKFLAGS etc.
(CHECKFLAGS is for sparse) because
you apparently mimick compiler flags in kernel-doc.



BTW, kernel-doc is invoked from Documentation/Makefile too.

Do we need to pass the same flags to both of them?



> johannes






--
Best Regards
Masahiro Yamada