Things are looking pretty good over on the Python 3 branch, thanks to the great work done at CTPUG, Sage Days 29 and elsewhere, and it probably bears some discussion about where we'd like to go from here. I'm thinking we should probably do another 1.0.x release with all of the bugfixes we currently have on that branch. Then, the question is when to start treating the py3 branch as master. The only downside on Python 2.x (that I'm aware of) is the dropped support for Python < 2.6. There are still a number of reasons (mainly due to external library dependencies) that one may not to run matplotlib on Python 3.x, but that's a separate question, IMHO. The status of the Py3 work is here: https://github.com/matplotlib/matplotlib-py3/wiki I would love a Mac user to build and run regression tests on both Python 2.x and Python 3.x, and run some interactive examples, particularly the mac os backend, as I haven't been able to test that personally. Thoughts? Mike -- Michael Droettboom Science Software Branch Space Telescope Science Institute Baltimore, Maryland, USA
Hi Mike, On Tue, Mar 29, 2011 at 12:52 PM, Michael Droettboom <md...@st...> wrote: > Things are looking pretty good over on the Python 3 branch, thanks to > the great work done at CTPUG, Sage Days 29 and elsewhere, and it > probably bears some discussion about where we'd like to go from here. > > I'm thinking we should probably do another 1.0.x release with all of the > bugfixes we currently have on that branch. > > Then, the question is when to start treating the py3 branch as master. > The only downside on Python 2.x (that I'm aware of) is the dropped > support for Python < 2.6. There are still a number of reasons (mainly > due to external library dependencies) that one may not to run matplotlib > on Python 3.x, but that's a separate question, IMHO. > > The status of the Py3 work is here: > > https://github.com/matplotlib/matplotlib-py3/wiki > > I would love a Mac user to build and run regression tests on both Python > 2.x and Python 3.x, and run some interactive examples, particularly the > mac os backend, as I haven't been able to test that personally. > > Thoughts? How about the following: 1) Cut a new v1.0.x release, update the documentation at the sourceforge site, and announce on all the mailing lists that development has moved to github. 2) Announce that the v1.1.x series will be the last to support <=python-2.5, and that a v1.1.x feature freeze is about to go into effect. Once devs have the features committed that they will need with <=python-2.5, create the v1.1.x maintenance branch and cut the 1.1.0 release. 3) Merge matplotlib-py3/master into matplotlib/master, champagne and doughnuts 4?) Make v1.2.0alpha1 binaries available so bleeding edgers can help identify issues on Windows (and maybe OS X). Darren
On Tue, Mar 29, 2011 at 12:52 PM, Michael Droettboom <md...@st...> wrote: > I would love a Mac user to build and run regression tests on both Python > 2.x and Python 3.x, and run some interactive examples, particularly the > mac os backend, as I haven't been able to test that personally. I'm not able to get the cocoaagg backend working with v1.0.1 and python-2.7. When I set the backend to cocoaagg in matplotlibrc, start "ipython -pylab", and do "plot([1,2])", I get a figure window with no axes, and an error message: 2011年03月30日 11:48:30.149 Python[18435:903] Inconsistent set of values to create NSBitmapImageRep I get the same results if I install from the latest sources or if I install using macports. This is with OS X 10.6.6 and pyobjc-2.2. I am unable to install and test with pyobjc-2.3 from pypi, several of the subpackages fail to install, including pyobjc-framework-Cocoa. Darren