Hi all,
The world of wxAgg and versions of wxPython has been a big pain for a while. I've been poking around trying to get MPL working with wxPython on OS-X with wxPython2.8, and it's been far too much work.
However, Ken McIvor has apparently solved the problem. Unfortunately, he committed the changes on Feb 22, and it looks like the 0.90.0 release was done on Feb 6. Darn.
My understanding of the changes is:
There is now code in backend_wxagg.py that uses methods to create and manipulate wx.Bitmaps that is new to wxPython2.8. This should make the old accelerator code obsolete.
At runtime, the code determines if you are running >2.8, and if so, uses the new code. If not, it reverts to the old code, which checks for the accelerator, uses it if it's there, and uses the pure-python non-accelerated code if not.
So -- this should allow one to build a single executable that will work well with wxPython 2.6, 2.8 and future wxPythons.
Personally, I think we should just drop the accelerator altogether. If someone needs better performance, let them use wxPython2.8 ( I think there is some room for improvement in the pre-2.8 python code too -- you could build the wxImage from a buffer, rather than a string, saving one data copy).
If the accelerator is dropped, then the whole build process gets easier.
If people want to keep the accelerator, I wrote some code for setup_ext.py that checks what version of wx you have installed, and finds the wx-config that matches it. It could be adapted a bit to try to build only a wxPython2.6 version, and not try to build a 2.8 version, which fails at this point.
So, with a bit more testing and tweaking, we could get this all settled.
What's the consensus on keeping the accelerator?
Is it worth putting out a point release with these fixes?
-Chris
···
--
Christopher Barker, Ph.D.
Oceanographer
Emergency Response Division
NOAA/NOS/OR&R (206) 526-6959 voice
7600 Sand Point Way NE (206) 526-6329 fax
Seattle, WA 98115 (206) 526-6317 main reception
Chris.Barker@...236...