26_numerics/c99_classification_macros_c.cc

Gabriel Dos Reis gdr@integrable-solutions.net
Sat Jun 14 23:26:00 GMT 2003


Jim Wilson <wilson@tuliptree.org> writes:

[...]

| Unforutately, it is lacking details about exactly why the change was
| made, but we can guess that we were getting cascading errors, and thus
| a single dg-error line was not sufficient.  We would get cascading
| errors if fpclassify was defined as a macro in math.h, and indeed it
| is on my system.  I offer no opinion on whether math.h or libstdc++ is
| wrong.  I haven't bothered to check any standards.

Actually it is libstdc++ which is wrong [some may say it is the
standard which is wrong :-)].  We are supposed to do our outmost to
provide a <math.h> that is clean.  That is only possible if we control
the C version of <math.h>, something we don't on non glibc systems.
As I understood it, Benjamin worked with Ulrich to put better support
in glibc-2.3.

-- Gaby



More information about the Gcc mailing list