Re: __dead functions

["Greg A. Woods; Planix, Inc." <woods%planix.ca@localhost>]

On 18-Jul-08, at 10:51 AM, Alexander Shishkin wrote:
> Or maybe an assertion can be placed right before __NOTREACHED?

An assertion is a run-time failure, not a compile time failure.

Even if it were true that this function were never to return, I'm  
still not sure an assertion would be the best thing.  That would  
depend entirely on the context.

> /*
> + * Explicitly specify that a certain point in the codepath should not
> + * be reached and indicate so to gcc.
> + */
> +#if __GNUC_PREREQ__(4, 3)
> +#define __NOTREACHED for (;;) continue
> +#else
> +#define __NOTREACHED
> +#endif

I think inventing extra redundant identifiers like this one, even if  
prefixed by "__" is unnecessary.

Indeed gcc doesn't understand a NOTREACHED comment, that's a lint  
directive of course and it seems, unfortunately, that lint has fallen  
out of favour in NetBSD.  However you can tell gcc a function doesn't  
return with "__attribute__ ((noreturn))" attached to a prototype.

<sys/cdefs.h> already defines an empty "__attribute__" macro for non- 
GCC compilers.

--
                                        Greg A. Woods; Planix, Inc.
                                        <woods%planix.ca@localhost>