On Sat, Aug 26, 2023 at 10:12 AM Masahiro Yamada <masahiroy@xxxxxxxxxx> wrote:
>
> On Thu, Aug 24, 2023 at 7:50 PM Sergey Senozhatsky
> <senozhatsky@xxxxxxxxxxxx> wrote:
> >
> > On (23/08/24 10:00), Masahiro Yamada wrote:
> >
> > > For the MFD_RK808 case particularly,
> > > I believe Kconfig showed MFD_RK8XX_I2C
> > > as a new option.
> >
> > I think there were some other unmet dependencies for MFD_RK8XX_I2C and
> > I don't think the new config was shown. But even if it was, we still
> > would have no idea that this meant "MFD_RK808 is not available anymore
> > and the corresponding code won't get compiled". So the "this is not
> > recognized anymore" is still needed and is quite helpful.
> >
> > Would you be OK with "list missing" being a warning (not a terminal
> > condition)?
>
>
> I am fine with implementing both.
>
> But, I'd like to implement them as separate options.
> (one option for warn unknown symbols,
> another for for turning warnings into errors)
>
>
>
> As I replied to Tomasz, I am considering about
> env variables vs command line options associated with W= option.
With a little more thought, the command line option approach
would require more code changes and efforts.
So, I am OK with adding new env variables.
Could you add two env variables?
I think
the first two hunks (show warnings for symbols not found in Kconfig)
-> KCONFIG_WARN_UNKNOWN_SYMBOLS
the last hunk (turn warnings into errors)
-> KCONFIG_WERROR
(You can suggest a better naming if you have, but I guess
KCONFIG_WARN_* will be consistent in case
more warning requests come up.)
--
Best Regards
Masahiro Yamada
