Re: gawk infinity issues

[Andrew J. Schorr]

On Fri, Jan 06, 2006 at 10:38:05AM -0500, Andrew J. Schorr wrote:
> So my question
> is why not rewrite force_number to call strtod unless the string looks
> like one of the hexadecimal values that we want to exclude?

In other words, what's wrong with the attached patch?  It seems to fix
the problem for me on all 3 of my platforms (linux x86 32 & 64 bits,
and solaris 8 sparc 32 bits).  With the patch, the attached test script
notfinite.awk does not produce any output other than the sqrt warning:

   $ ./gawk -f test/notfinite.awk
   gawk: test/notfinite.awk:17: warning: sqrt: called with negative argument -1

But without the patch, I see the following on my linux boxes:

   $ ./gawk -f test/notfinite.awk
   gawk: test/notfinite.awk:17: warning: sqrt: called with negative argument -1
   (sprintf(%f,nan) = nan) != (sprintf(%f,"nan"+0) = 0.000000)
   (sprintf(%f,nan) = nan) != (sprintf(%f,"-nan"+0) = 0.000000)
   (sprintf(%f,inf) = inf) != (sprintf(%f,"inf"+0) = 0.000000)
   (sprintf(%f,-inf) = -inf) != (sprintf(%f,"-inf"+0) = 0.000000)

And under solaris, the unpatched version gives:

   $ ./gawk -f test/notfinite.awk                    
   gawk: test/notfinite.awk:17: warning: sqrt: called with negative argument -1
   (sprintf(%f,NaN) = NaN) != (sprintf(%f,"NaN"+0) = 0.000000)
   (sprintf(%f,Inf) = Inf) != (sprintf(%f,"Inf"+0) = 0.000000)

And I think the patch should allow the removal of gawk_strtod.

Regards,
Andy

P.S. I'm a bit concerned about locale-related issues that might affect
this test:

   if ((*cp == '0') && ((*(cp+1) == 'x') || (*(cp+1) == 'X'))) {

On the other hand, isnondecimal seems to contain similar logic...

force_number.patch
Description: Text document

notfinite.awk
Description: Text document