Why Special float.h Files? (Was: Re: PATCH: XFAIL c99-float-1.c...)

[Jeffrey Oldham <oldham at codesourcery dot com>]

Joseph S. Myers writes:
> Why isn't the testsuite getting GCC's <float.h> included instead of the
> system one?  That seems to me to be a bug - where GCC provides a header
> required of freestanding implementations, testcases should be able to
> include it.  The point of this testcase is to remind target maintainers
> that they need to update their <float.h>, by failing until they do so.

I will back out the patch.

Richard Henderson writes:
> Incorrect.  GCC provides this header.  However, x86 is the
> only version of float.h that has been updated.

Assuming gcc will construct its own float.h header files rather than
relying on the header files provided by the operating system,

1) How do we ensure that gcc's float.h are consistent with the
   underlying system since the underlying system can change without
   warning us?  Do I just guess correct values?

2) Who uses these float.h header files?  gcc only?  gcc users?

3) Why are a proliferation of gcc/config/float-*.h files appearing
   when the fixinc mechanism exists to modify existing system files
   into a form that gcc needs?  Using this mechanism will help ensure
   consistency between the system-provided files and what gcc needs.
   It will also eliminate the need to define tens of float-*.h files.

4) Where is documentation explaining these float-*.h files and which
   gcc/config.gcc items to change?  For example, only some of the
   i[34567]86-*-* entries specify "float_format=i386" but I cannot
   discern why some cases are omitted.

Thanks for your help,
Jeffrey D. Oldham
oldham@codesourcery.com