WX/WXagg backend add code that zooms properly on a Mac with a Retina display

[briantoby]

This uses a slower method for drawing the zoom rubberband box on all Macs (perhaps not needed in Phoenix?) because I don't know a way to detect if a Retina display is present.

Note that the added self.canvas.draw() call in the FigureManagerWx.show() method comes from the version in the Canopy dist for 1.5.0.

[jenshnielsen]

Thanks, I will try to take a look later. Last time I checked it was not fixed in Phoenix I think it is a bug in Wx and not the python bindings

[RobinD42]

I'll review the change today as well.

[RobinD42]

It looks good to me, other than a style mismatch (self.RetinaFix should probably be self.retinaFix to match the existing style in the module).

In addition, it looks like there is an undocumented method in wx that can be called to help determine if retina is currently in use for a window or a DC. It's not wrapped by wxPython yet but I'll get it added and try to get at least a preview build done soon so it can be tested. In the meantime you may want to add a reminder comment in the code that when window.GetContentScaleFactor() and/or dc.GetContentScaleFactor() are added then they can be used to determine if the retina workround is needed. (Or perhaps we can actually use it to fix the overlay problem...)

[briantoby]

Hmmm, travis failure makes no sense. Passed that before and error messages seem unconnected to the one file I changed. What gives?

[WeatherGod]

It is a spurious LaTeX error that we get sometimes on Travis. I restarted that particular test.

[WeatherGod]

Tests now pass. I don't have a retina display, so I can't do much else. I'll leave this up to one of the other devs to finalize.

[jenshnielsen]

backported to 1.5.x as 911f349