Build hangs on psMatrixVectorArithmetic

[Paul Price]

Building psLib using the autoconf stuff is not successful --- it hangs on
psMatrixVectorArithmetic.

price@mithrandir:/home/mithrandir/price/psLib4/pslib-1.4>./configure
--prefix=/home/mithrandir/price/psLib4/pslib-1.4/
[...]
price@mithrandir:/home/mithrandir/price/psLib4/pslib-1.4>unalias make
(remove parallelism through -j to see what's causing it)
price@mithrandir:/home/mithrandir/price/psLib4/pslib-1.4>make
[...]

gcc -DHAVE_CONFIG_H -I. -I. -I../../src -I../../src/astronomy

-I../../src/collections -I../../src/fileUtils -I../../src/image
-I../../src/sysUtils -g -O2 -Wall -std=c99 -D_GNU_SOURCE -I/usr/localinclude
-I/usr/localinclude/libxml2 -MT psMatrixVectorArithmetic.lo -MD -MP -MF
.deps/psMatrixVectorArithmetic.Tpo -c psMatrixVectorArithmetic.c -fPIC -DPIC -o
.libs/psMatrixVectorArithmetic.o

The build sits at this point a very long time. I've let it run for several
minutes. "top" reveals that "cc1" is chewing a lot of processor:
/usr/lib/gcc-lib/i386-redhat-linux/3.2.3/cc1 -lang-c -I. -I. -I../../src
-I../../src/astronomy -I../../src/c[snipped by window]

The same happens on Opteron box alala.

[robert.desonia@…]

Actually, that file takes a very long time to compile given it is basically
'macro hell'. It can take up to a minute on our machines here; try to be
patient and see if it finishes.

[robert.desonia@…]

I'll look into splitting this file into two: one for BinaryOp and one for
UnaryOp. That might be more paletable.

[Paul Price]

OK, I see --- it's the optimisation. If I just wait on our Opteron (alala), the
build (with default O2 optimisation) takes a total of 4m25s (4m of which is
psMatrixVectorArithmetic). If I set CFLAGS='-g -O0', then it takes only 16s.

Splitting it into two might be better, since dual processor machines with "make
-j2" could work in parallel.

It might also help to make a note in the build docs to help impatient people
like me.