In Python 3.10, a change[1] was implemented where extension functions
that take integer arguments will no longer silently accept non-integer
arguments (e.g., floats) that can only be converted to integers with a
loss of precision. This PR fixes most of these issues in the pure-Python
classes and demos by explicitly converting the parameters to int before
passing them to wxWidgets. There is loss of precision, but this was
happening before (automatically) anyway as most wxWidgets DeviceContext
functions operate using integers.
Additionally, the PR fixes a few sizing issues, mostly with SpinCtrls being
too small on GTK3.
This is an example of the relevant exception:
Traceback (most recent call last):
File "/usr/lib64/python3.10/site-packages/wx/lib/agw/pygauge.py", line 355, in OnPaint
r.width = w
TypeError: 'float' object cannot be interpreted as an integer
Fixes#2038.
[1] https://bugs.python.org/issue37999
Fixes Python 3.10 and should be faster not having to download & install
Python and pip.
This also removes Python 3.6 from macOS CI - the Azure hosted version of
Python 3.6 has a problem with linker flags (it is setting -stack_size) and
can't compile Python extensions. Python 3.6 is going to be EOL in a month or
so, so it seems fine to just remove it from CI.
wx.Region no longer selects all non transparent pixels when a wxBitmap is passed to the constructor. In order to get the region made properly a color bust also be supplied. As luck would have it the way the docking guide images were made we are able to do this because there is only a single color that we needed to use.
wx.lib.agw.aui was never converted to run on Python 3 and as a result the division operator forces a conversion of the used values to a float. Floats cannot be used in functions like `range`. This caused quite number of things to not function properly and produce tracebacks.
The other thing is when wx.lib.agw.aui was written pixels could not be rendered using coordinates that were floats so now passing floats could cause rendering alignment problems because the layout of the various bits were done so in a manner that would have the alignment correct using integers.
