Backport of #3687 to 1.4.x

[mdboom]

No description provided.

[mwaskom]

Is there a timeline for a 1.4.3 release with this fix?

[tacaswell]

The current target date is Feb 1 with an RC in early/mid January.

[mwaskom]

Any hope for something sooner? This bug completely breaks seaborn.

[danielballan]

+1 for considering an urgent 1.4.3 release to fix this.

Also, would it make sense to test downstream libraries like seaborn in mpl's test suite so that mpl would at least alerted to major breakages like this?

[mwaskom]

+1 for considering an urgent 1.4.3 release to fix this.

FWIW I have cut point release with a hacky workaround for this bug, so it's no longer quite so urgent.

Also, would it make sense to test downstream libraries like seaborn in mpl's test suite so that mpl would at least alerted to major breakages like this?

It actually probably makes a bit more sense for me to have one of the Travis builds run on matplotlib's github master.

[danielballan]

OK, great, I discovered 0.5.1 after I commented. Thanks, @mwaskom.

[tacaswell]

Given the discussion of v2.0 coming soon and off-line discussions with
@mdboom and @efiring about doing bug-fix releases more often it probably
will happen sooner rather than later (mostly contingent on my finding the
bandwidth to do the release).

I don't think mpl should get into the business of testing down-stream
libraries, that way lies madness.

In this case we also had a failure in the automated tests (there is a
tolerance on them to deal with slight variations in font rendering which
allowed the test that should have caught this to pass) so it probably
would not have helped in this case anyway.

On Mon Nov 24 2014 at 10:50:31 AM Dan Allan notifications@github.com
wrote:

+1 for considering an urgent 1.4.3 release to fix this.

Also, would it make sense to test downstream libraries like seaborn in
mpl's test suite so that mpl would at least alerted to major breakages like
this?

—
Reply to this email directly or view it on GitHub
#3699 (comment)
.