Re: loslib.c warning

[Glenn Maynard <glenn@...>]

On Mon, Sep 25, 2006 at 10:59:57AM +0100, David Jones wrote:
> >   exit(luaL_optint(L, 1, EXIT_SUCCESS));
> >   return 0;  /* to avoid warnings */
> >
> >I'd recommend removing this; it's introducing a legitimate warning in
> >order to fix a buggy one.
> 
> It entirely depends on what compiler one is using and what warnings  
> you have enabled.

I disagree.  The warning that this is (presumably) meant to fix is "no
return value in function returning non-void".  Any compiler implementing
that warning should also be implementing __attribute__((noreturn)) or
equivalent.

glibc does it: extern void exit (int __status) __THROW __attribute__ ((__noreturn__));
and VC does it (since VC6): _CRTIMP __declspec(noreturn) void __cdecl exit(int);

The "unreachable code" warning may have pros and cons, but handling
"noreturn" to avoid spurious return value warnings is basic.  That's
why it seems like a backwards trade.  (Even if it's a minor one; the
only reason I bothered mentioning it was because I found it funny that
a line with the comment "to avoid warnings" was the cause of a warning.)

I wonder what compiler this was actually for.  CodeWarrior?

> #define ASSERT(x) if(1) {} else do { sizeof(x); } while(0)
>
> This way the compiler still checks the condition for syntax, etc, but
> no code gets generated.  There's lots of unreachable and dead code.

I havn't seen a compiler that would trigger that warning on this code;
that would be a bit over-sensitive.

-- 
Glenn Maynard