Re: [PATCH] Make WARN_ON/WARN_ON_ONCE no-ops when CONFIG_BUG is off

[Benjamin Herrenschmidt <benh@kernel.crashing.org>]

On Fri, 2007-12-14 at 21:27 +0800, Herbert Xu wrote:
> Hi:
> 
> [PATCH] Make WARN_ON/WARN_ON_ONCE no-ops when CONFIG_BUG is off
> 
> The description of CONFIG_BUG clearly states that both BUG and
> WARN_ON may be skipped.  However, our actual implementation still
> checks the condition on WARN_ON if it's used as part of an if
> statement or such.
> 
> This patch makes it return 0 after evaluating the expression
> if CONFIG_BUG is disabled.  This is consistent with the spirit
> of the CONFIG_BUG option.
> 
> The same change is made to WARN_ON_ONCE.
> 
> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

That's something I've actually never quite liked... the fact that we
evaluate the expression anyway. I'm pretty happy with -not- evaluating
the expression when CONFIG_BUG is on most of the time since whatever is
in there is purely here for the sake of the BUG/WARN test.

I understand why some people may want it the other way around, but I
personally find it a very bad idea in the first place to write a normal
statement part of the program as

	BUG_ON(do_something());

It's way clearer to me I believe to write:

	rc = do_something();
	BUG_ON(rc);

And thus, when I write:

	BUG_ON(do_sanity_check());

I'm actually execting the function call to disappear when CONFIG_BUG
is turned off...

Cheers,
Ben.