Regression: odd rendering of zordered areas on twinx axes in 1.2 (release) versus 1.1 #1807

phobson opened this Issue Mar 4, 2013 · 2 comments


None yet
3 participants

phobson commented Mar 4, 2013

(Note: I'm going to follow up on this with a git bisect to track down the issue.)

In trying to implement horizon charts, I came across an odd regression.

The code that demostrates the error is here:

I brought this up on StackOverflow ( and the answerer used my original code in the SO question and matplotlib 1.1 to produce this correct, but poorly formatted image:


But as shown in the SO link, with matplotlib 1.2,. and a recent build from master, I get:


As an aside, an implementation that works on both versions and is perhaps a bit more elegant can be founder here:

If appropriate, I'd like to work this into a PR, although it may be more appropriate for pandas.


mdboom commented Mar 8, 2013

I can't reproduce the correct result on 1.1.0 or 1.1.1. What specific version did this work on? There are some complicated interactions between invert_yaxis and twinx going on here, but it would first be helpful to establish a baseline that's working. Is this on the Agg backend, or some other backend?


pelson commented May 22, 2013

I can't reproduce the correct result on 1.1.0 or 1.1.1.

I can reproduce the correct result... with master. @phobson - I'm going to close this, but if you can't get the correct result with master, we can re-open and investigate.

For what it's worth, I was able to get the failing result with v1.2.0 and v1.2.1 but not with v1.2.x - but then we have fixed a lot of backend_agg bugs recently, I guess this is the fruition of that 😄

pelson closed this May 22, 2013

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment