Bugzilla Bug 102872

I created an ebuild pygtk-2.6.1-r1.ebuild which uses numarray instead of 
numeric. 
See: http://bugs.gentoo.org/attachment.cgi?id=66154 
 
But I don't knoiw about possible side-effects, so could somebody check this one 
and tell me if we need to patch scigraphica or can just go on and let pygtk 
using numarray instead of numeric? 
 
Waiting for comments. 
 
Best regards Jan 

Reproducible: Always
Steps to Reproduce:
1.
2.
3.

Why is this a blocker? Also, scigraphica depends on numeric, not numarray. From
which bug is the attachment? Please, clarify the whole thing. Thanks.

OK, I should give more info, you are correct.  
  
This one blocks scigraphica-2.1.0 new ebuilds are available at: 
http://bugs.gentoo.org/show_bug.cgi?id=93095 
 
The problem is the following: 
Adrian (author of sg) changed dependency of sg from numeric to numarray. So I'm 
standing on the line: 
If I try to install sg using my new ebuilds I get numarray (dependency of sg) 
and numeric ('cause of IUSE of pygtk; I don't know the correct reason, but it 
seems to be default on x86) 
 
Now scigraphica give strange results if both numeric *and* numarray are 
installed. 
 
So there are two possible solutions: 
 
1. pygtk can use numarray instead of numeric (thats the ebuild I provided in 
the mentioned bug-report) 
2. We need to patch the installation routine of sg and hope, there are no 
side-effects if both numeric *and* numarray are installed. 
 
Thanks for clearification! ;) 
 
Best regards Jan 
 
I'll reopen till we get the "real" solution. 

OK, I added the block for Bug 93095 and CCed the maintainers of scigraphica;
hopefully we'll be able to find some solution. Thanks for explanation.

Any new developments about this?

(In reply to comment #2)

> Now scigraphica give strange results if both numeric *and* numarray are 
> installed. 

Could you specify which are those strange results? I am trying to work on the
scigraphica ebuilds in bug #93095. I tried building pygtk disabling numeric but
did not see any change in scigraphica behaviour.

This bug doesn't make sense. Besides, package.masked for gnome-1 removal.

WONTFIX.