sinf and cosf (was: RE: xscale uclibc softfloat & gdb : SIGILL)

[John Bowler]

From: Eric BENARD / Free
>I had to switch from uClibc 0.9.27 to 0.9.28 because of cosf & sinf
functions
>which are not in 0.9.27 when using softfloat.

The problems I encountered with sinf and cosf in .27 were because gcc 3.4.4
automagically converts cos(float) and sin(float) to cosf/sinf with certain
(almost any) optimisation levels.

The fix for that is -fno-builtin-sin -fno-builtin-cos to gcc when compiling,
good to know they are there in 0.9.28 though.

John Bowler <jbowler at acm.org>