Two or three initial observations on the fp prining patch

[Jerry Quinn]

I was looking in the i686-pc-linux-gnu/config.cache for potentially
useful stuff.  There's already vars defining sizeof several types:

ac_cv_sizeof_int=${ac_cv_sizeof_int=4}
ac_cv_sizeof_long_double=${ac_cv_sizeof_long_double=12}

Following the trail into the autoconf docs, there's AC_CHECK_SIZEOF.
Sizeof long double could be used to select between ieee quad and
extended.  Also, there appears to be an endian test:

ac_cv_c_bigendian=${ac_cv_c_bigendian=no}

With several of these, I assume it should be possible to select the
appropriate types to use.  I don't know if it's possible to have a
bigendian machine with littleendian floating point or vice versa.

This stuff still doesn't help with the VAX or IBM issues.  Do these
arch's have long double, anyone?

Jerry