#7 setup on RHEL fails due to gsl_error issue

[Anonymous]

I have installed gsl v1.9, but setup of pygsl-0.9.4 during the build stage with this error:

gcc -pthread -fno-strict-aliasing -DNDEBUG -g -O3 -Wall -Wstrict-prototypes -fPIC -DSWIG_COBJECT_TYPES=1 -DDEBUG=1 -DNUMERIC=1 -DPYGSL_GSL_MAJOR_VERSION=1 -DPYGSL_GSL_MINOR_VERSION=9 -UNDEBUG -I/kali1/agarg/pipeline/v9.0sm/photpipe/Cfiles/include -IInclude -I. -I/kali1/agarg/pipeline/v9.0sm/photpipe/python/include/python2.3 -c src/init/initmodule.c -o build/temp.linux-i686-2.3/src/init/initmodule.o
In file included from src/init/block_helpers.c:256,
from src/init/initmodule.c:51:
src/init/block_helpers_numarray.ic:33:89: macro "gsl_error" passed 4 arguments, but takes just 0
src/init/block_helpers_numarray.ic:72:48: macro "gsl_error" passed 4 arguments, but takes just 0
In file included from src/init/block_helpers.c:256,
from src/init/initmodule.c:51:
src/init/block_helpers_numarray.ic: In function `PyGSL_nummarray_convert':
src/init/block_helpers_numarray.ic:72: warning: statement with no effect
src/init/block_helpers_numarray.ic: In function `PyGSL_Copy_Array':
src/init/block_helpers_numarray.ic:33: warning: statement with no effect
error: command 'gcc' failed with exit status 1

To be extra sure I don't have some buggy version of gsl, I looked at gsl_errno.h and it looks fine. Is gsl_error redefined somewhere?

[Pierre Schnizer]

Yes, gsl_error was redefined in
Include/pygsl/error_helpers.h
<snip>
#ifndef PyGSL_ERROR_HELPER_H
#define PyGSL_ERROR_HELPER_H 1
#include <pygsl/intern.h>
#include <pygsl/utils.h>
#include <gsl/gsl_errno.h>
#include <pygsl/errorno.h>

#undef GSL_ERROR
#undef GSL_ERROR_VAL
#undef GSL_ERROR_NULL
#define gsl_error()
</snip>

This was done for debugging purposes... just delete the lines
#undef GSL_ERROR
#undef GSL_ERROR_VAL
#undef GSL_ERROR_NULL
#define gsl_error()

and everything should run okay. I will take this code out from the next release.

SIncerely yours

[Pierre Schnizer]

Thank you for this bug report. Please indicate if you are still interested in numarray and will not using
numpy in the future. Its also of interest as I was thinking of ceasing the support of numarray as a package. But if there are users the support can be left in. But be aware that the ufuncs (special and complex functions) are not available for numarray.

The bug is currently limited (as far as I can see in numarray) I will fix it in CVS and in the next releases.

Sincerely yours
Pierre