klibc - Aug 2010 - ISO C99 inline, again

Hi,

I just saw this wasn?t merged for 1.5.20 (at least the stdio.h
part, which I submitted somewhere earlier).

I did a grep for "extern.*inline" and touched all functions,
but this should be tested first (the klibc/compiler.h and
ctype.h ones, at least), as I was not 100% clear on which
behaviour exactly was *intended* in the first place.

The general idea is:
if __GNUC_STDC_INLINE__ use ?static inline? else ?extern inline?,
and add __attribute__((__always_inline__)) if the compiler is new
enough ? which means to inline and not add a function version of
the code.

bye,
//mirabilos
-- 
I believe no one can invent an algorithm. One just happens to hit upon it
when God enlightens him. Or only God invents algorithms, we merely copy them.
If you don't believe in God, just consider God as Nature if you won't
deny
existence.		-- Coywolf Qi Hunt
-------------- next part --------------

On 08/29/2010 08:11 AM, Thorsten Glaser wrote:
> Hi,
> 
> I just saw this wasn?t merged for 1.5.20 (at least the stdio.h
> part, which I submitted somewhere earlier).
> 
> I did a grep for "extern.*inline" and touched all functions,
> but this should be tested first (the klibc/compiler.h and
> ctype.h ones, at least), as I was not 100% clear on which
> behaviour exactly was *intended* in the first place.
> 
> The general idea is:
> if __GNUC_STDC_INLINE__ use ?static inline? else ?extern inline?,
> and add __attribute__((__always_inline__)) if the compiler is new
> enough ? which means to inline and not add a function version of
> the code.
> 
klibc is written in the same style as the Linux kernel, which is
currently GNU style for inlining; I am not going to support more than
one style for reasons of maintainability.

Furthermore, your patch actually *removes* a fix by which we enforce
GNU-style even if it is not the default, so it's in fact doubly wrong.

	-hpa