python 2.6 support - jaunty

This is odd. While I can't find where spectlib tries to import numpy, something in its dependency graph does. I'm not certain this is a specto bug, but there's certainly a bug somewhere!

Did you try that using the main specto bazaar branch? It seems to me
that fixing the other python 2.6 issues may have made this vanish, at
least I haven't experienced this particular problem.

solved by:

sudo apt-get install numpy

please add python-numpy to dependencies

received error while trying to use specto -> message about numpy module... will try the fix above.

sudo apt-get install python-numpy

solved the problem for me..

Might I ask wtf is this numpy module I've never heard about before as a
dev, though? :) it used to work without it.

numerical python?

http://sourceforge.net/projects/numpy

On Fri, Apr 10, 2009 at 2:49 PM, Jean-François Fortin Tam <
<email address hidden>> wrote:

> Might I ask wtf is this numpy module I've never heard about before as a
> dev, though? :) it used to work without it.
>
> --
> python 2.6 support - jaunty
> https://bugs.launchpad.net/bugs/354909
> You received this bug notification because you are a direct subscriber
> of the bug.
>

--
Luke Whitmore

UK Landline: +442030511409

More info might be needed on this issue upstream:
http://code.google.com/p/specto/issues/detail?id=216

Setting 'New' because I can reproduce this bug.

Please test the package available in my PPA:

http://ppa.launchpad.net/quadrispro/testing/ubuntu

I can confirm this issue and the suggested workaround.

Igor Gomes

Alessio, what changes did you make to your PPA version? perhaps you could provide your patch on http://code.google.com/p/specto/issues/detail?id=216 ?

For what it's worth, the last time I tried swapping hashlib and md5, this caused serious problems (web watches not working at all).

As you can see in the debdiff, I only added a dependency on python-numpy.

-Depends: ${python:Depends}, python-gtk2, python-glade2, python-gnome2, python-gnome2-extras, zenity, gconf2, python-notify, python-dbus, librsvg2-common, gnome-icon-theme
+Depends: ${python:Depends}, python-gtk2, python-glade2, python-gnome2, python-gnome2-extras, zenity, gconf2, python-notify, python-dbus, librsvg2-common, gnome-icon-theme, python-numpy

We should also fix the deprecated modules. They are easy enough.

Uploaded to my testing PPA

Everything works fine.

Uploaded, sitting in the queue.

Reviewed the diff, looks appropriate to me. ACK.

This bug was fixed in the package specto - 0.2.2-3ubuntu1

---------------
specto (0.2.2-3ubuntu1) jaunty; urgency=low

  * Add missing Depends on python-numpy (LP: #354909).
  * Add 10_fix_python26_warnings.patch patch to fix deprecation warnings.

 -- Alessio Treglia <email address hidden> Mon, 20 Apr 2009 07:56:11 +0200

Hold on a sec, Specto does not call the numpy module directly in its code, afaik, PyGTK is the one calling it. Shouldn't PyGTK be the package that gets its dependencies fixed? Surely we can't be the only application with which this happens?

IIRC, there was a decision made for PyGTK not to carry the numpy depends
because numpy wouldn't fit on CD and it was thought the relevant functions
were only rarely used. This isn't the "right" fix, but as a workaround for
this release, it will do.

pygtk uses numpy for exactly one method. the reason not to have it as a dependency are space constraints on the CD (numpy and dependencies would add 10MB). So probably the current python-gtk package should be renamed into python-gtk-lite and an empty python-gtk package depending on python-gtk-lite and python-numpy. Then all packages going on the CD need to depend on python-gtk-lite.

I suspect that changing the users of this one method to add that dependency is a lot more practical than renaming python-gtk.

Jaunty has reached his EOL on 23 October 2010.

I'm marking this bug as incomplete, please test on a new ubuntu version and set the status again to "new" if you think the bug is not fixed.

Just an FYI to anyone who comes across this, I've located the exact call that depends on numpy. It is quite difficult to find this (as numpy is only used indirectly), so I've decided to comment despite the age of this issue.

It is line 327 of spectlib/notifier.py

    for row in icon.get_pixels_array():

You can duplicate the segmentation fault by running the following:

import gtk
icon = gtk.icon_theme_get_default().load_icon('dialog-information', 64,0)
print(icon.get_pixels_array())