calling sincosf on subnormal argument is wrong

[Joseph S. Myers]

On Thu, 16 Aug 2012, Dmitrieva Liubov wrote:

> > Since the
> > bugs arising from calling sincos are currently latent, no need to file
> > them in Bugzilla before sending a patch
> 
> Ok, I will file them after a patch with optimized sincos.

What I meant was: you can send patches to *fix* the uses of sincos without 
needing to file associated bugs; a patch known to expose latent bugs as 
testsuite failures is unlikely to be approved.  And then propose optimized 
sincos once the latent bugs (that optimized sincos would expose) are 
fixed.  I don't think these latent bugs should be that hard to fix, and as 
you've found there are already relevant testcases in the testsuite for 
them.

-- 
Joseph S. Myers
joseph@codesourcery.com