[libgfortran,patch] Two small fixes

[FX <fxcoudert at gmail dot com>]

1. With clock.diff, I make sure we undefine the TCK macro before redefining it; this silences a warning I'm getting on Mac OS 10.6.

2. With literal.diff, we introduce a macro system similar to stdint.h's INT64_C macros to define literal floating-point constants of a given kind. The macro appends the right suffix (f for float, nothing for double, l for long double, and soon to come... q for __float128). This provides an indirection layer between kinds and floating-point types, which we will need later on for __float128 support. It's not the most fundamental change ever, but I think it's nicer.


Currently regtesting on x86_64-linux. OK to commit once regtesting finishes?

FX

Attachment: clock.diff
Description: Binary data

Attachment: clock.ChangeLog
Description: Binary data

Attachment: literal.diff
Description: Binary data

Attachment: literal.ChangeLog
Description: Binary data