Subject: Re: CVS commit: src/sys/arch/i386/stand/genprom
To: Tom Spindler <dogcow@babymeat.com>
From: Perry E. Metzger <perry@piermont.com>
List: tech-misc
Date: 12/27/2007 09:45:23
Tom Spindler <dogcow@babymeat.com> writes:
>> Alan Barrett <apb@cequrux.com> writes:
>> > Since bail() doesn't return, it seems to me that the __dead marker was
>> > correct.  The proper fix is not to remove the __dead, but to change the
>> > build so that __dead is defined.
>> 
>> Compiles with high WARNS levels would now fail...
>
> Have you tried compiling with a high WARNS level? No errors are
> thrown, since <bsd.hostprog.mk> doesn't do anything with WARNS.

Not the point. There is no reason to be less concerned about bugs in
programs (maintained by NetBSD) are used as part of the build
process. cc -Wall etc. is a useful tool for that...

> In any case: the only time a warning would be emitted is if you
> (rather inexplicably) turned on -Wmissing-noreturn - which isn't
> enabled by any other -W flags.

Not true. I had some fun with this during test compiles while
replacing various __attribute__ uses.

I agree with your later comment that we should discuss this on
tech-misc, so I'll take your reply there...

-- 
Perry E. Metzger		perry@piermont.com