[PATCH 019/171] lib: Add a Kconfig for SPL_GENERATE_ACPI_TABLE

[Tom Rini]

On Mon, Jan 30, 2023 at 04:18:28PM +0100, Mark Kettenis wrote:
> > Date: Mon, 30 Jan 2023 09:52:52 -0500
> > From: Tom Rini <trini at konsulko.com>
> > 
> > On Mon, Jan 30, 2023 at 03:50:23PM +0100, Heinrich Schuchardt wrote:
> > > On 1/30/23 15:40, Simon Glass wrote:
> > > > This is implicitly used in the source and seems useful, so add it.
> > > 
> > > Please, provide a clear description how this will be used.
> > 
> > To follow up on what I just sent, Simon seems to be taking a pass at
> > adding to Kconfig many / most of the options that aren't available today
> > but are implied by the usage of CONFIG_IS_ENABLED(FOO). So, this is
> > another case where IS_ENABLED(CONFIG_FOO) should be used.
> 
> Hi Tom, Simon,
> 
> I fear we have a documentation/communication/naming issue here.
> Before Simon pointed out an issue to me in the s5p serial driver my
> mential picture of this was:
> 
>   #if -> CONFIG_IS_ENABLED(FOO)
> 
>   if () -> IS_ENABLED(CONFIG_FOO)
> 
> possibly because that's the pattern you see if you grep the source
> code (IS_ENABLED(CONFIG_FOO) is rarely used with preprocessor
> conditionals).  And that makes me think I'm not the only developer
> that got this wrong.
> 
> After I realized that the purpose of CONFIG_IS_ENABLED(FOO) is to
> cover TPL, SPL and U-Boot proper with a single condition it was fairly
> obvious to me that there are many, many cases where
> IS_ENABLED(CONFIG_FOO) should be used instead.
> 
> And I fear Simon's series only makes it harder to fix that.  Right now
> you can easily see that CONFIG_IS_ENABLED(FOO) is probably the wrong
> construction since no CONFIG_TPL_FOO and CONFIG_SPL_FOO Kconfig
> symbols exist.

This is why I'm not sure that moving away from #ifdef for the most part
is helpful. Simon has a plan to make things so that we only ever use one
form of the macro, because SPL/TPl/VPL are handled differently. I'm not
sure and need to read it all over again once it's re-posted as I have
some concerns. I'm still not convinced that a lot of cases of if
(MACRO(SYMBOL)) { ... } actually read any better than #ifdef SYMBOL ...
#endif.

-- 
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 659 bytes
Desc: not available
URL: <https://lists.denx.de/pipermail/u-boot/attachments/20230130/99808ca1/attachment.sig>