A.G.Grozin@inp.nsk.su reports problem with redhat 7.0:
1. Yesterday, I tried to install Maxima 5.6. It seems that there were no
error messages during make (there were some warnings while making gcl
2.4.0), but many self-tests fail. Also, I get the same behaviour as
reported recently in sci.math.symbolic: integrate(x,x); yields
x^2*false^2/2, integrate(sin(x),x); produces division by zero. This is on
I believe this is related to the buggy gcc version 2.96 released with
redhat/cygnus. Basically it is badly broken on -O4 optimize and it is
hard to see how they could have released it. I have made some patches
to gcl's configure and to the compiler invocation code, to not use
-O4 when this compiler version is there. Please
I have patched the configure
in gcl to detect that buggy version. Note that maxima is now compiled
on -O4 by default, and the above configure patch and patch to the
compiler invocation in gcl, will shut off -O4 for this compiler.
Basically configure if it detects this compiler version should add
a #define BROKEN_O4_OPT, and the broken_o4_opt feature gets added
to the lisp, so that maxima will not get compiled with -O4.
So you should take the latest cvs/gcl or gcl daily snapshot to get
those changes if you are using redhat. I dont have a redhat system
here myself, so if someone could please check this out.
Thanks
william