svn commit: r215237 - head/lib/msun/src

[David Schultz]

On Mon, Nov 15, 2010, Bruce Evans wrote:
> On Mon, 15 Nov 2010, Alexander Best wrote:
> 
> >also it seems one of the regression tests in tools/regression/lib/msun 
> >fails
> >(assert tripped): test-lrint.

The test is failing due to an actual deficiency in gcc; you need
to disable the built-in versions of lrint and friends for it to
pass.  I considered working around that in the test (as in
test-conj.c) and tricking the compiler into using the library
versions, but that really just hides the problem.

> I get this on amd64, and lots more failures starting with csqrt() on
> i386.

Yes, some work is needed to deal with i386's special precision
issues.  As you know, it's often hard to get gcc to even produce
the right code on i386, and I didn't spend an inordinate amount of
effort trying to get tests for long doubles to pass on that
platform.