Dave Korn dave.korn@artimi.com
Sun Jun 26 16:05:00 GMT 2005

----Original Message----
>From: Christopher Faylor
>Sent: 16 June 2005 17:49

> On Thu, Jun 16, 2005 at 05:39:22PM +0100, Dave Korn wrote:
>> ----Original Message----
>>> From: Christopher Faylor
>>> Sent: 16 June 2005 16:06
>>> On Thu, Jun 16, 2005 at 03:36:46PM +0100, Dave Korn wrote:
>>>> weekend.  My tuits are also being taken up currently with the newlib
>>>> lrint problem...
>>> I would like to get a 1.5.18 out soon, and lrint is one of the things
>>> I'd like to get into cygwin.  Doesn't gcc provide built-in versions of
>>> some of "lrint" style functions?  Would it make sense to use those
>>> somehow?
>> It does, and we could use them, or we could adopt the mingw inline
>> asms, and drop newlib's version altogether.  I don't think I'm going to
>> have any time to put to this until saturday, but I think that I
>> understand the source of the problem: the representation of zero in FP
>> is a special case, yet the code treats it as if it had a valid mantissa
>> and exponent.
> I was looking at mingw's versions, too.  I wouldn't mind just pulling
> them over into cygwin.  That would continue a trend that I've been
> following of moving as much stuff away from newlib as possible.  We
> would need non-inline versions, for backwards compatibility, though.
> I'm going to be out of town for most of next week, so it looks like a
> 1.5.18 release won't be until Saturday at the earliest.
> cgf

Posted a fix for the newlib problem at 


It would still make sense to adopt the mingw inline asms as well, and we
might as well pinch their fenv.h and add functions like fegetround and
fesetround, but that's a second patch!

Can't think of a witty .sigline today....

More information about the Cygwin-developers mailing list