On 07/24/2013 11:36 AM, Michiel de Hoon wrote:
Hi Michael,

> Thanks.  I will address these issues before final. 

Thanks.

> None of this is new -- we've gone through 5 release candidates over many weeks.  Why is this just coming to light now?

The output of running "python setup.py build" passes by quickly, and people (including myself) did not notice that the MacOSX backend was not being compiled.
And then if an older version of matplotlib is already installed, everything seems to be working fine.
I noticed this problem today only because I was installing matplotlib on a new Macbook.

Ah, that makes sense.  Hopefully as part of doing some automated Mac builds we can be testing clean installs etc. and catch this kind of thing sooner in the future.

Mike


Best,
-Michiel.


From: Michael Droettboom <mdroe@stsci.edu>
To: Michiel de Hoon <mjldehoon@yahoo.com>
Cc: Benjamin Root <ben.root@ou.edu>; "matplotlib-devel@lists.sourceforge.net" <matplotlib-devel@lists.sourceforge.net>
Sent: Wednesday, July 24, 2013 10:36 PM
Subject: Re: [matplotlib-devel] 1.3.0rc5 tagged

Thanks.  I will address these issues before final. 

None of this is new -- we've gone through 5 release candidates over many weeks.  Why is this just coming to light now?

Mike

On 07/23/2013 11:16 PM, Michiel de Hoon wrote:
I noticed that this release candidate will always skip the Mac OS X backend,
as the .check() method of BackendMacOSX in setupext.py returns None. Adding
    return "darwin"
or some other string solves the issue.
Also, if I am not mistaken, the files under lib/matplotlib/backends/Matplotlib.nib (as returned by the get_package_data() method of BackendMacOSX in setupext.py) are needed by the cocoaagg backend, not by the MacOSX backend.

Best,
-Michiel.
_______________________________________________