compiling grass5.0beta7 on linux

[Markus Neteler]

Hi all,

my apologies for my yesterday's INFINITY statement.
It seems to be a problem related to rather all platforms
beside my personal platform (?).

But you can be shure - GRASS 5 shall become compliant to 
all compilers/platforms, not limited to gcc compiler.

I hope we find a general solution soon.

Sorry for any inconvenience

 Markus Neteler


On Mon, Apr 24, 2000 at 06:56:59PM -0700, Russell L. Carter wrote:
> 
> Hi all,
> 
> %Markus Neteler wrote:
> %> Hi there,
> %> 
> %> INFINITY is defined in
> %> 
> %> /usr/include/math.h:    INFINITY        representation of the infinity value of type 'float'
> %> 
> %> If r.cost is not compiling on your machine, I assume your compiler
> %> is not installed properly. The declarations in src/raster/r.cost seems
> %> to be o.k.
> %> 
> %> Kind regards
> %> 
> %>  Markus Neteler
> %
> %Hi Markus, 
> %
> %sorry to insist on that, but on my Red Hat Linux 5.2, glibc 2.0.7, 
> %INFINITY is not defined in /usr/include/math.h or elswhere.
> %
> %On IRIX 6.2 INFINITY is not defined too and compilation of r.cost stops
> %too. 
> 
> It is also not in any versions of FreeBSD, include 5.0-current.
> Nor is isfinite ().  I copied over the macro and inline function
> defs from RH-6.1 into the one file that needed it.
> 
> %Please do not assume that all Unix systems are the same or compatible
> %with some standards Linux or glibc introduced a very short time ago. I
> %still hope that GRASS will continue support for other Unix platforms
> %than Linux. 
> %
> %A portable program should use HUGE_VAL for a undefined infinite which is
> %defined in <math.h> and in <limits.h>. HUGE_VAL should be compatible
> %with systems that use IEEE floating point format. Another way is the
> %function isinf(double X), but this is BSD and is not portable to systems
> %that strictly confine to SYSV. 
> %
> %I don't know if the code in main.c is correct as INFINITY is defined as
> %a float but null_cost is a double variable. Other possible defines:
> %FLT_MAX and DBL_MAX from <limits.h> contain the biggest value a float
>  ^^^^^^^^^^^^^^^^^^^
> These are available on lots of systems... #include <float.h>
> 
> There is also that "#ifdef linux" in a .l file which fixes yylineno, 
> needed for BSD too.  Thankfully somebody left a ~ file around, a 
> diff revealed the fix.
> 
> Other than that... a clean build.  
> 
> Thanks!
> Russell
> 
> 
> %resp. a double variable can hold. One could use FLT_MAX instead of
> %INFINITY. 
> %
> %Please contact me if you whish me to investigate this further. I can
> %provide a patch.
> %
> %cu 
> %
> %Andreas 
> %
> %-- 
> %Andreas Lange, 65187 Wiesbaden, Germany, Tel. +49 611 807850
> %Andreas.Lange at Rhein-Main.de, A.C.Lange at GMX.net
> %
> 
>