Fix ctan, ctanh of subnormals in round-upwards mode (bug 14328)

[Adhemerval Zanella]

On 07/08/2012 05:17 PM, Joseph S. Myers wrote:

> Actually I noted that sqrt (epsilon) is really what's relevant here (just 
> not convenient to write generically), so you might as well use DBL_EPSILON 
> or 0xp1-53L or similar.
>
> I'd say such constants are better written as hex floats, 0x1p-106L in this 
> case (any floating-point constant where you intend an exact bit pattern 
> for a particular floating-point format is better written as a hex float 
> unless there is a short exact decimal expansion for it).
>
Thanks for the review, I'll rewrite the epsilon constant as hexa format, but I think
it is best to keep the ldbl-128ibm implementation close to the one you already committed,
instead of fix the spurious overflow using another logic. Don't you think?


-- 
Adhemerval Zanella Netto
  Software Engineer
  Linux Technology Center Brazil
  Toolchain / GLIBC on Power Architecture
  azanella@linux.vnet.ibm.com / azanella@br.ibm.com
  +55 61 8642-9890