On Sat, Dec 29, 2001 at 11:44:28PM +0200, Jarkko Hietaniemi wrote:
> and these
>
> pp.c:2711: warning: passing arg 2 of `modf' from incompatible pointer type
>
> mean that something in the
>
> 2711: (void)Perl_modf(value, &value);
>
> Perl_modf() logic in perl.h has gone wrong. I hope you can investigate.
our logic about what to do if ${OS} doesn't have a modf for long doubles is
bust. I'm not sure what the best thing to do is:
Temporary variables?
Bar Configure for long doubles unless a long double modf is found?
[not such a preposterous idea, considering that modf is used in several
places]
Obviously finding one for Sparc (IIRC Solaris also didn't have one) would be
a good thing. But we still need to fix the logic.
Can we roll our own modf? [Would it be numerically stable for extreme
values] Unpack's use of modf for checksums can be replaced with fmod, IIRC.
Do Sparc Solaris and Sparc Linux have a long double modf?
[and is the lack of modf an indication that the CPU doesn't have an FP
instruction for modf at long double precision?]
On Sat, Dec 29, 2001 at 04:56:33PM -0500, Michael G Schwern wrote:
> I'm just a monkey with a lot of hammers, I wouldn't know where to
> start looking. :(
Hammers are good. Select the right hammer for your job though. :-)
Claw hammers, ball pane hammers, club hammers, sledge hammers are all the
types I can think of offhand.
[Oh, and "Birmingham screwdriver" which is a hammer by another name]
Topic? What's that? Oh. sorry. I thought that this was the London.pm list. :-)
Nicholas Clark
