The Fastest Route to C99?

[Joseph S. Myers]

On Sat, 7 Dec 2002, Scott Robert Ladd wrote:

> I'm a number-cruncher by trade, and the C99 floating-point extensions appeal
> to me. Given my interest in seeing gcc become fully C99-compliant, I want to
> contribute.
> 
> The question is: Contribute WHERE?
> 
> Not being a compiler-writing guru, I want to be very cautious about mucking
> with code generation. Is anyone specific coordinating C99 compliance?

No.  I haven't had time for doing much on it lately.  I think the table in
<http://gcc.gnu.org/c99status.html> is accurate, but the notes below it
haven't been updated for changes in how <float.h> is handled.

(Of course I don't know whether anyone has contracted any of the companies 
doing GCC development to complete the C99 implementation.)

> And while C++ library support resides in the gcc project (i.e., libstdc++ is
> distributed with gcc), C support is in glibc (which is distributed separate
> from the compiler). Many unfinished aspects of C99 lie in the realm of
> glibc, a project that appears a tad cloistered, likely due to the central
> nature of glibc (i.e., upgrading it tends to break existing systems).

The glibc status is documented in the CONFORMANCE file distributed with
glibc, though that may be out of date (I haven't checked it against glibc
2.3).  For an outside view of glibc's status, and criticisms thereof, see
recent discussions on the libc-alpha list.

> I'm somewhat at a loss as to how to proceed. For example, I have expertise
> in complex numerics and IEEE 754 math. I'd like some pointers as to where I
> might best expend my efforts.

Ask Stephen Moshier for the latest version of his patches towards the C99
pragmas.  There was a lot more that needed to be done (e.g. getting the
pragma status attached to expressions as they pass through trees and RTL)  
but they should be a useful starting point, including some testcases.

-- 
Joseph S. Myers
jsm28@cam.ac.uk