moving toolmanager initialization up before toolbar

[fariza]

PR Summary

When running the toolmanager example with the tk backend we get the following error

09:40 $ python3 examples/user_interfaces/toolmanager_sgskip.py 
Traceback (most recent call last):
  File "examples/user_interfaces/toolmanager_sgskip.py", line 79, in <module>
    fig = plt.figure()
  File "/home/federico.ariza/workspace/matplotlib/lib/matplotlib/pyplot.py", line 521, in figure
    **kwargs)
  File "/home/federico.ariza/workspace/matplotlib/lib/matplotlib/backend_bases.py", line 3214, in new_figure_manager
    return cls.new_figure_manager_given_figure(num, fig)
  File "/home/federico.ariza/workspace/matplotlib/lib/matplotlib/backends/_backend_tk.py", line 1025, in new_figure_manager_given_figure
    manager = cls.FigureManager(canvas, num, window)
  File "/home/federico.ariza/workspace/matplotlib/lib/matplotlib/backends/_backend_tk.py", line 515, in __init__
    self.toolbar = self._get_toolbar()
  File "/home/federico.ariza/workspace/matplotlib/lib/matplotlib/backends/_backend_tk.py", line 535, in _get_toolbar
    toolbar = ToolbarTk(self.toolmanager, self.window)
  File "/home/federico.ariza/workspace/matplotlib/lib/matplotlib/backends/_backend_tk.py", line 816, in __init__
    ToolContainerBase.__init__(self, toolmanager)
  File "/home/federico.ariza/workspace/matplotlib/lib/matplotlib/backend_bases.py", line 3030, in __init__
    self.toolmanager.toolmanager_connect('tool_removed_event',
AttributeError: 'NoneType' object has no attribute 'toolmanager_connect'

In PR #11559 the tk toolbar initialization was moved up (packaging problem).
When working with the toolmanager, the toolmanager has to be initialized before the toolbar.
This fix just moves up the toolmanager initialization

PR Checklist

• Has Pytest style unit tests
• Code is PEP 8 compliant
• New features are documented, with examples if plot related
• Documentation is sphinx and numpydoc compliant
• Added an entry to doc/users/next_whats_new/ if major new feature (follow instructions in README.rst there)
• Documented in doc/api/api_changes.rst if API changed in a backward-incompatible way

[jklymak]

Why wasn't this caught by tests?

[anntzer]

Because test coverage of backend_tools is close to zero.

[fariza]

Are we testing the backends? this is related to specific TK implementation

[anntzer]

Yes: https://github.com/matplotlib/matplotlib/blob/master/lib/matplotlib/tests/test_backends_interactive.py

[jklymak]

There are various test_backendFoo in the test directory, but I don't know the details. But that need not hold up this PR if noone has a good idea how to add a test for this...

[anntzer]

Someone "just" needs to write them, the model in test_backends_interactive.py should be enough... (but we don't need to hold this PR up for that)

[tacaswell]

Thanks @fariza

Please try not to push branches to the main matplotlib repo!

[fariza]

Ups sorry, I didn't notice I wasn't in my dev desktop. And without thinking I pushed to "origin" and of course it pointed to main repo.

…

On Wed, Jul 18, 2018, 1:10 PM Thomas A Caswell, ***@***.***> wrote: Thanks @fariza <https://github.com/fariza> Please try not to push branches to the main matplotlib repo! — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#11694 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABa86SRuCB1wWtMTMk8zfXhQQbvwXbPoks5uH2wRgaJpZM4VUtB5> .

[tacaswell]

@fariza No worries!