in reply to Avoiding perl's Atof when assigning floating point values

More about Windows at a later date

As I mentioned in my initial post, the patch to numeric.c needs some adjustment for Windows because of https://sourceforge.net/p/mingw-w64/bugs/711 and https://sourceforge.net/p/mingw-w64/bugs/725.
Eventually, I discovered that there was a fairly simple and effective workaround to this problem. It's just a matter of replacing one of the Perl_strtod calls with a call to __mingw_strtold. This adjusted numeric.c patch that portably caters for both Windows and Linux can be found at http://www.sisyphusion.tk/scratch/numeric.c.diff.txt.
With the patch in place, mingw-w64 built perl-5.28.0 exhibits the same dramatic improvement in its assignment of floating point values as was seen on my Ubuntu box. (See my original post for details.)

I'm not sure what effect (if any) this patch will have on a perl-5.28.0 that was built by a Microsoft compiler - but that's something I'll try to determine over the coming week.

Cheers,
Rob
  • Comment on Re: Avoiding perl's Atof when assigning floating point values (MS Windows)