Created attachment 422338 [details] Patch EPEL5 numpy.spec to add python26 support as well. Hi, Here is a patch to the current EPEL5 .spec file to build a python26-numpy package as well from the same .spec file. Of course it may easier to do a completely separate an up to date numpy. Steve. In fact although the package builds fine there are is an ERROR and FAIL from the python26 version of numpy but I think its good to record the intention here. ERROR: Tests the min/max functions with explicit outputs ---------------------------------------------------------------------- Traceback (most recent call last): File "/var/tmp/numpy-1.2.1-3.el5-root-steve/usr/lib64/python2.6/site-packages/numpy/ma/tests/test_core.py", line 699, in test_minmax_funcs_with_output result = npfunc(xm,axis=0,out=nout) File "/var/tmp/numpy-1.2.1-3.el5-root-steve/usr/lib64/python2.6/site-packages/numpy/core/fromnumeric.py", line 1569, in amin return amin(axis, out) File "/var/tmp/numpy-1.2.1-3.el5-root-steve/usr/lib64/python2.6/site-packages/numpy/ma/core.py", line 3119, in min np.putmask(out, newmask, np.nan) ValueError: cannot convert float NaN to integer ====================================================================== FAIL: test_umath.TestComplexFunctions.test_against_cmath ---------------------------------------------------------------------- Traceback (most recent call last): File "/usr/lib/python2.6/site-packages/nose/case.py", line 183, in runTest self.test(*self.arg) File "/var/tmp/numpy-1.2.1-3.el5-root-steve/usr/lib64/python2.6/site-packages/numpy/core/tests/test_umath.py", line 268, in test_against_cmath assert abs(a - b) < atol, "%s %s: %s; cmath: %s"%(fname,p,a,b) AssertionError: arcsin 2: (1.57079632679-1.31695789692j); cmath: (1.57079632679+1.31695789692j)
Created attachment 422339 [details] Byte compiles with path option. This is simple script that makes the byte compiling a little more flexible to cope with this dual python. It's needed by the patched .spec file.
Created attachment 426411 [details] python26-numpy package for EPEL5. I think this is better solution. This is a python26-numpy.spec file that builds only a python26 variant of numpy version 1.4.1. It's largely based on the F14 one again. I'll probably submit this as a new package soonish but you may have comment. The tests all pass fine now with this. I'm fairly confident that this will not interfere with the existing numpy in EPEL5 and I realise that is most important. Steve.
Sorry on the next one I won't include: /usr/include/numpy -> /usr/lib64/python2.6/site-packages/numpy/core/include/numpy/ but /usr/include/numpy26 -> /usr/lib64/python2.6/site-packages/numpy/core/include/numpy/
I agree, I think that would be better, but I think it might be best to provide the same version of numpy for the two python versions. If you submit a review and want to maintain it, that's fine, otherwise I can submit and maintain, if you want to do the review. Your call.
I'd prefer the new version, apart from anything else it does not just work with the python2.4 version, see comment #1. Given we have a super new python 2.6 we may as well take new stuff where possible given there is no interaction between the two things. I realize this is newer than rawhide. And we will be stuck with this version for a few years I guess. I'll submit the package shortly with this bug blocking on it. Of course subsequent co-ownership makes sense if you want to ask. Steve.
Thank you Steve, your spec builds fine. But I'd like to have an option to prune lapack and Fortran dependencies using build options
I'll close this bug since this was an RFE to create a python26-numpy package based on the existing EPEL5 version of numpy. This is now obsoleted by the ongoing package review of python26-numpy in bug #619355.