fortran-using components need to be split out

Bug #309215 reported by Steve Langasek
6
Affects Status Importance Assigned to Milestone
pygtk (Ubuntu)
Fix Released
Undecided
Unassigned
Jaunty
Fix Released
Undecided
Unassigned
python-numpy (Ubuntu)
Won't Fix
Low
Unassigned
Jaunty
Won't Fix
High
Unassigned

Bug Description

Binary package hint: python-numpy

With the migration of pygtk from python-numeric to python-numpy in jaunty, we've had a net loss of roughly 6MB of disk space on the CDs; python-numeric was split into python-numeric and python-numeric-ext binary packages, but python-numpy now pulls in blas, lapack, and libgfortran into the CD dependencies, in addition to being significantly larger than python-numeric was due to all the code being shipped in the same package.

This has effectively caused all of the language packs to be shifted off the CDs for alpha-2, and even then the amd64 desktop image is still oversized. For alpha-3, we should really try to reclaim this space by splitting the python-numpy package.

Related branches

Steve Langasek (vorlon)
Changed in python-numpy:
importance: Undecided → High
milestone: none → jaunty-alpha-3
status: New → Confirmed
Revision history for this message
Matthias Klose (doko) wrote :

for alpha-2 we just did remove the build-dependency on lapack and blas. not optimal. nor sure if we can fix it for alpha-3.

Revision history for this message
Matthias Klose (doko) wrote :

splitting out a sub-package is ugly. there are plans to include numpy into the python core, but it looks like these are stalled (http://numpy.scipy.org/array_interface.shtml).

the original request was to get the CD size down, and the only package depending on numpy is pygtk. pygtk does use the multiarray module for two functions (one of them is deprecated): GdkPixbuf.pixel_array and GdkPixbuf.get_pixels_array. Both functions are not used in main, afaics, so it is difficult to see why this interface was added and how it is used.

maybe just raise an ImportError if the module cannot be imported, with the message append "please install python-numpy"?

Revision history for this message
Matthias Klose (doko) wrote :

that means, still b-d on python-numpy to compile in the runtime check, and move the dependency to a suggests or recommends.

Revision history for this message
Steve Langasek (vorlon) wrote :

This is still an open question for jaunty, but there's nothing that blocks alpha-3 here. Retargetting to alpha-6.

Changed in python-numpy:
milestone: jaunty-alpha-3 → jaunty-alpha-6
Revision history for this message
Andrew Straw (astraw) wrote :

The page at http://numpy.scipy.org/array_interface.shtml is out of date. This work became PEP 3118, and is now part of Python 2.6 and 3.0: http://www.python.org/dev/peps/pep-3118/

Also, as far as I know, there are no plans to put Numpy into Python core.

As far as pygtk, once it supports PEP 3118, it's doubtful that it will need to depend on numpy. I guess that's unlikely to happen before Jaunty, however.

Revision history for this message
Launchpad Janitor (janitor) wrote :

This bug was fixed in the package pygtk - 2.13.0-2ubuntu5

---------------
pygtk (2.13.0-2ubuntu5) jaunty; urgency=low

  * Remove the dependency on python-numpy, make it a suggestion. Avoids
    python-numpy on the CD. Nothing in main does use this feature, plus
    an exception at runtime is thrown if the extension cannot be found.
    Packages using this feature must depend on python-numpy explicitely.
    See the bug report for the details. LP: #309215.
  * Build for python2.6.
  * debian/rules (clean): Remove .pyc files.

 -- Matthias Klose <email address hidden> Sun, 22 Feb 2009 00:03:00 +0000

Changed in pygtk:
status: New → Fix Released
Matthias Klose (doko)
Changed in python-numpy:
status: Confirmed → Won't Fix
Revision history for this message
Matthias Klose (doko) wrote :

python-gtk2 doesn't depend anymore on python-numpy. The size of numpy is still an issue, but I currently don't see a sane way to make this smaller. Building it twice, once with the included lapack_lite, and with the available libraries, and then diverting, doesn't sound like the right choice.

Changed in python-numpy:
importance: High → Low
milestone: jaunty-alpha-6 → none
status: Confirmed → Triaged
Revision history for this message
Steve Langasek (vorlon) wrote :

marking this as 'wontfix' overall for python-numpy, it appears we have a correct solution for pygtk now.

Changed in python-numpy:
status: Triaged → Won't Fix
Martin Pitt (pitti)
Changed in python-numpy:
milestone: jaunty-alpha-6 → none
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.