Re: [9fans] testing and p9fmt

Re: [9fans] testing and p9fmt

[Russ Cox]

There are a few other things to be aware of.
First, fmtdef.h says

	typedef unsigned long uvlong;

which should be

	typedef unsigned long long uvlong;

on systems that support it.

As for nan, I believe that this

	static uvlong uvnan    = 0x7FF0000000000001;

	double
	nan(double *d)
	{
		USED(d);
		return *(double*)&uvnan;
	}

will work on systems that need it, as long as
they use 64-bit doubles.

Russ

Re: [9fans] testing and p9fmt

[William Josephson]

On Thu, Dec 05, 2002 at 11:48:54PM -0500, Russ Cox wrote:
> There are a few other things to be aware of.

You also want to define _POSIX_SOURCE (e.g.
by using -posix on FreeBSD) or remove the
ushort and uint typedefs on BSD.

  -WJ

[9fans] testing and p9fmt

[Scott Schwartz]

This message is testing a change to our mailing list software.

But as long as you're listening, the p9fmt library that Russ mentioned
recently needs a tweak to compile on Solaris.

In strtod.c

#ifdef __sun
#include <sunmath.h>
#define nan(x) quiet_nan(0)
#endif

In Makefile,

CFLAGS=-xO5 -xipo=1
LDLIBS=/opt/SUNWspro/lib/libsunmath.so
LDFLAGS=-R/opt/SUNWspro/lib

[The -xipo flag turns on optimization in the linker, which is
more Plan 9-like. :)]