[Ecls-list] ecl on Solaris/sparc?

[Raymond Toy]

Michael Abshoff wrote:
 > Raymond Toy wrote:
 > Shouldn't you add "-std=c99" to CFLAGS? Solaris is "correct" about that
 > since it is a C99 feature IIRC :)

Another problem.  math.h defines isfinite to a call it __builtin_finite, 
but gcc doesn't know what that is so it won't link.  Perhaps the gcc 
build isn't quite right.

But a Solaris 8 box doesn't have isfinite at all because it's not C99.  
My solution to this was to #define isfinite(x) finite(x) where finite is 
available.  May not be quite right since there's no version for 
single-float, but I think the cast for single-float infinity to 
double-float should produce an infinity.

I just put this in ecl.h.

With this change, and using some prebuilt gmp, I am able to build a 
working ecl on Solaris 8 at work.

 >
 > Sounds like a gas issue I had seen before. The Sun Freeware tools ship
 > an ancient binutils. I "fixed" it by removing the "-g" flag since the
 > problem was dwarf2 debugging symbols, but I might be very wrong here.

I'm using gcc that uses Sun as and ld, and not binutils.  It's now 
building on a different Solaris 8 box using the gmp included with ecl.  
I think it will work. :-)  But the version of gcc on the Solaris 10 box 
uses binutils and that does have the problem.

I was able (once!?!?) to compile maxima with this ecl and pass maxima 
test suite, so I think it's working pretty well. :-)

Ray