newlib math bug (fpclassify)

["Joseph E. Vornehm, Jr." <joev at mitre dot org>]

I've already posted this info to the newlib list along with a patch. 
Someone suggested I re-posting it here, since it affects development of 
scientific/mathematical software using Cygwin.  Apologies if this is 
repeat information for anyone; this seemed to be a more appropriate 
forum than, say, cygwin-patches.

newlib contains a bug in its math libraries that causes fmax() and 
fmin() not to handle negative numbers properly.  I presume several other 
math functions are affected, as well.  Specifically, the __fpclassifyf 
and __fpclassifyd functions incorrectly report negative numbers as NaNs.

newlib 1.11.0 is affected; I presume prior versions are affected, as 
well.  Version 1.3.22 of the Cygwin DLL is affected.  The newlib cvsweb 
interface shows that the patch has been applied to the CVS tree.

Joe Vornehm



--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Problem reports:       http://cygwin.com/problems.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/