[coreboot] kbuild troubles

[ron minnich]

On Sun, Oct 11, 2009 at 4:36 PM, Peter Stuge <peter at stuge.se> wrote:

>> if (CONFIG_FOO) {
>> }
>>
>> which fit in more nicely than the separate layer of preprocessing
>> statements
>
> Why is this more nice? Since it's not really common I would be
> surprised (and slightly annoyed) to not find CONFIG_FOO declared in
> any source code. OK, the prefix helps, but it still looks a little
> strange to me.

It's way more nice.

First off, if your compiler optimizes, then it can just optimize this
out, getting the same effect as #ifdef.

Second, it really removes the problems of #iifdef. Sometimes people
forget to #define something and the #ifdef fails, which is not what
they might have expected :-). Sometimes something might be defined to
0, meaning don't use it, but people test it with #ifdef, oops. I have
seen all these scenarios.

Third, it makes the CONFIG variables consistent. It's a pain when some
config variables have variable (e.g. base addresses and baud rates and
so on) and others do not.

The whole #ifdef thing happened 30 years ago. It's obsolete. We can
now set and test cpp variables, and we might as well.

Rob Pike has articulated this argument far better than I can, and I'm
on vacation in Santa Fe today :-), so I'll only say this: let's leave
it with cpp variables having values. I don't want to see us go back to
#ifdef.

thanks

ron