M_LOG2_E vs. M_LN2 cleanup patch

Howland Craig D (Craig) howland@LGSInnovations.com
Fri Dec 3 15:36:00 GMT 2010

     Yes, that would be better.  (It's in there way too many times, and
missed line 532.)  Adjusted patch to do that attached.
     (By the way, there is no practical difference in the values, beause
double does not have enough precision to get there.  Intel 80-bit long
double probably would not notice, either.  But is is best to define it
once only, as you suggest.)
     Thanks for the sharp eyes.

-----Original Message-----
From: newlib-owner@sourceware.org [mailto:newlib-owner@sourceware.org]
On Behalf Of Corinna Vinschen
Sent: Friday, December 03, 2010 10:06 AM
To: newlib@sourceware.org
Subject: Re: M_LOG2_E vs. M_LN2 cleanup patch

On Dec  3 08:49, Howland Craig D (Craig) wrote:
> Jeff:
>      The math.h part of the patch appears to have not worked, as
> is unchanged in CVS.  Here's that part of the patch, re-generated
> 				Craig

I was going to apply the patch, but I was wondering if we really want
to stick to this discrepancy:

line 35:  #define _M_LN2  0.693147180559945309417

line 532: #define M_LN2   0.69314718055994530942

Shouldn't we better define M_LN2 in terms of _M_LN2:

line 532: #define M_LN2 _M_LN2



Corinna Vinschen
Cygwin Project Co-Leader
Red Hat
-------------- next part --------------
A non-text attachment was scrubbed...
Name: mathh.patch
Type: application/octet-stream
Size: 1900 bytes
Desc: mathh.patch
URL: <http://sourceware.org/pipermail/newlib/attachments/20101203/6ef1d7a7/attachment.obj>

More information about the Newlib mailing list