TST: drop py2.6 & py3.3 testing #5215
Conversation
|
This is a pre-req for getting #4762 merged. |
|
@matthew-brett @blink1073 Who do I ask nicely to get a pandas 3.5 wheel built? |
|
On it. |
|
Now available. |
|
If this passes I will squash these to 1 or 2 commits. |
|
Thanks! On Thu, Oct 8, 2015 at 4:08 PM Steven Silvester notifications@github.com
|
As discussed on the mailing list, going forward we will be dropping support for python 2.6 and python 3.3. This allows mpl to use ordered dicts and is required for in-progress traitlet/serialization work. - moved minimum support numpy test to 2.7 - moved pandas optional test dep to 3.5 http://matplotlib.1069221.n5.nabble.com/Matplotlib-devel-supported-python-versions-tt46154.html
7c06e9e to
5beb7ab
Compare
| - python: 2.7 | ||
| env: MOCK=mock PANDAS=pandas | ||
| - python: 3.3 | ||
| env: MOCK=mock NUMPY=numpy==1.6 |
There was a problem hiding this comment.
Are there any plans to drop support for numpy 1.6? I think version 1.7 came out in early 2013, and supporting 1.6 seems to accumulate some preprocessor cruft.
There was a problem hiding this comment.
We should ask the numpy folks how many versions back they are supporting bug fixes for and don't go further back than that.
I am 👍 on dropping old versions of numpy, but that should be a separate discussion.
|
Also a prerequisite for #5295. I'm 👍 on merging this. This is marked for the 2.1 milestone, but if I understood the mailing list post correctly, we'll be dropping official support for Python 2.6 in 2.0, so I think it makes sense to stop testing with 2.6 already in 2.0. |
|
I'd want confirmation from @tacaswell that our understanding is correct that we'll drop Python 2.6 for matplotlib 2.0. It would be nice, because now we have a dependency that "new styles" --> depends on --> "new fonts" --> depends on --> "fixing font file caching" --> depends on --> "lru_cache which hasn't been backported to 2.6". So it will be hard to disentangle removing Python 2.6 support. |
|
I think the net result of the mailing list discussion was that we are going to drop 'official' support for 2.6 in 1.5 as we do not want commit to releasing a 2.6 compatible bug fix 1.5.1. No one volunteered to manage that and no one said that 2.6 is critical to them. We should probably back-port this to the 1.5.x branch as soon as we make it as well. |
|
Sure. Let's release 1.5.0 final and then immediately merge this and backport it to 1.5.x. |
|
This is blocking other PRs, might as well merge this now. I don't think we will have issues remembering to backport it tomorrow 👿 |
TST: drop py2.6 & py3.3 testing
|
👏 |
|
Backported to v2.0.x as 1b243ed |
TST: drop py2.6 & py3.3 testing
As discussed on the mailing list, going forward we will be dropping
support for python 2.6 and python 3.3.
This allows mpl to use ordered dicts and is required for in-progress
traitlet/serialization work.
http://matplotlib.1069221.n5.nabble.com/Matplotlib-devel-supported-python-versions-tt46154.html