Re: RFC/RFA: strtod() magnitude problem

[nick clifton <nickc at redhat dot com>]

Hi Corinna,

> > One point - this part of your patch:
> >
> > >   static _CONST double tinytens[] = { 1e-16, 1e-32, 1e-64, 1e-128,
> > > -		9007199254740992.e-256
> > > +		9007199254740992.*9007199254740992.e-256
> > >   		};
> >
> > appears to be a typo.  I have removed it in my revised patch, but
> > maybe it needs fixing ?
>
> I'm not sure.  I copied this from the NetBSD code verbatim:
>
> /* The factor of 2^106 in tinytens[4] helps us avoid setting the underflow */
> /* flag unnecessarily.  It leads to a song and dance at the end of strtod. */
> static CONST double tinytens[] = { 1e-16, 1e-32, 1e-64, 1e-128,
>                  9007199254740992.*9007199254740992.e-256
>                  };
>
> Note that not only the value itself is changed, the 2^53 in our version
> of the comment is replaced by a 2^106.  I just checked against OpenBSD,
> and their code contains the exact same change.
>
> So I think it was right.  Sure, even two upstream can be wrong, but still...

Ok - but please can we at least have a couple of spaces around the 
asterisk so that the it is more obvious that we are multiplying together 
two constants rather than trying to construct some kind of bizarre mega 
number.  Ie:
		
  9007199254740992. * 9007199254740992.e-256

Also - the last three entries in the tinytens[] array evaluate to 0 for 
a _DOUBLE_IS_32BITS target, and they produce a compile time warning 
message about that.  So I have attached a revised patch which explicitly 
stores 0.0 in these entries.  Is this version OK ?

Cheers
  Nick

Index: libc/stdlib/strtod.c
===================================================================
RCS file: /cvs/cvsfiles/devo/newlib/libc/stdlib/strtod.c,v
retrieving revision 1.31
diff -u -3 -p -r1.31 strtod.c
--- libc/stdlib/strtod.c        27 Jan 2013 12:46:12 -0000      1.31
+++ libc/stdlib/strtod.c        24 Apr 2013 08:23:58 -0000
@@ -130,8 +130,13 @@ THIS SOFTWARE.
 #undef tinytens
 /* The factor of 2^53 in tinytens[4] helps us avoid setting the 
underflow */
 /* flag unnecessarily.  It leads to a song and dance at the end of 
strtod. */
-static _CONST double tinytens[] = { 1e-16, 1e-32, 1e-64, 1e-128,
-               9007199254740992.e-256
+static _CONST double tinytens[] = { 1e-16, 1e-32,
+#ifdef _DOUBLE_IS_32BITS
+                                   0.0, 0.0, 0.0
+#else
+                                   1e-64, 1e-128,
+                       9007199254740992. * 9007199254740992.e-256
+#endif
                };
 #endif