[Bug 619355] Review Request: python26-numpy - A fast multidimensional array facility for Python

[bugzilla at redhat.com]

Please do not reply directly to this email. All additional
comments should be made in the comments box of this bug.


https://bugzilla.redhat.com/show_bug.cgi?id=619355

--- Comment #18 from Steve Traylen <steve.traylen at cern.ch> 2011-04-30 17:32:43 EDT ---
Hi Jerry,

Thanks for the comments so far.

New packages:

http://cern.ch/straylen/rpms/python26-numpy/python26-numpy-1.5.1-3.el5.src.rpm
http://cern.ch/straylen/rpms/python26-numpy/python26-numpy.spec


The easy ones:

* The isa tags indeed do nothing on EPEL5
* s/in in/be in/ done.
* Added -fno-strict-aliasing.

The other ones:

Concerning the liblapack in atlas and lapack:
I've been reading around the history of this, good news is we can drop
the requirement on lapack since indeed everything and more is also in atlas.
This also allows the specification of library locations to be dropped in 
the %build and %install sections as well. They were there from a time after
EPEL atlas when lapack in atlas was not including all of lapack, it's just not
needed
for EPEL5 versions of these packages.

The current situation with liblapack is definitely ugly and is
detailed in the bug #478856. Hopefully unrelated to this review I added a
comment
that there may now be a way out of this situation but the fix will come from 
atlas and lapack and will take time if it comes at all.


I'll look at the 
/usr/lib64/python2.6/site-packages/numpy/lib/_compiled_base.so
/lib64/libpthread.so.0
shortly.

I noticed you added a look more for the licensing, was it something in
particular, I will
look more anyway.

-- 
Configure bugmail: https://bugzilla.redhat.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.