@pwuertz Can you do the same for the pgf tests? I am having issues getting pgf to run on my machine.

I'm not sure I'm up to speed with the current development. What do you need? Check and recreate the reference images for the tests in the master and/or v2.x branch?

@pwuertz I did the last updates and am happy to do these to if it makes it any easier. Basically we just need to rerun the tests locally with the pgf dependencies installed and copy the new results to the baseline directory and remove the knowfailure on the pgf tests

Recreated baseline images with matplotlib master and removed knownfailure.

Commit is here: https://github.com/pwuertz/matplotlib/commit/36394382e9caaa0320ab1d5c30e401f4e1f1f93c

Do you want to integrate it into your own test update PR or do you want me to create 2 PRs for master and v2.x?

@QuLogic if you have band width please take this PR over

I'm trying this out, but more things failed on Travis than locally.

Or maybe it's because I didn't have xelatex installed, but unfortunately, now more things are failing locally than on Travis.

matplotlib.tests.test_axes.test_boxplot_rc_parameters passes by itself, but fails as a group. There might be a @cleanup decorator missing somewhere.

try clearing out your caches. that fixed a lot of issues for me, once.

On Thu, Sep 8, 2016 at 10:02 PM, Elliott Sales de Andrade <
notifications@github.com> wrote:

Or maybe it's because I didn't have xelatex installed, but unfortunately,
now more things are failing locally than on Travis.

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
#6757 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AARy-NMlMZkjRWhaRekILYv9j9RvKX-xks5qoL5HgaJpZM4JODzQ
.

if you change gs version and are not on master you have to clear .cache/matplotlib/test_*

I've rebased to latest v2.x and updated several test images due to various other changes occurring in the meantime.

There is still the question of matplotlib.tests.test_axes.test_boxplot_rc_parameters which works fine if run by itself, or as the first test, but fails if it's not. There don't appear to be any @cleanup missing, but there's clearly some setting that's falling through the cracks somewhere.

If this is fine, we should probably squash the commits to avoid any extra images taking up space.

And now we're ✅.

Now rebased to remove any redundant image changes.

@QuLogic Could your problem with matplotlib.tests.test_axes.test_boxplot_rc_parameters be related to _rc_test_bxp_helper not being private? I remember having had a similar issue of setting leaking through tests in #6035 that was due to such a reason: see @jenshnielsen last comment in #6035 .

@QuLogic Is this ready to go or does in need to be squashed?

@afvincent I am not sure to what you are referring in that issue, but I think the problem is fixed by 26774ca.

@tacaswell It is good to go from my side; I removed the redundant image changes.

@QuLogic If 26774ca made everything OK, then everything is fine :). For the record, I was referring to the issue of "matplotlib.tests.test_axes.test_boxplot_rc_parameters which works fine if run by itself, or as the first test, but fails if it's not." I had a similar issue with #6035 (a test that was OK if run alone, but not in the test suite) and @jenshnielsen found that it was related to the custom setup function I was using, which was not a test (i.e. not named test_*), was not private either. It was like the function was leaking some context to the ones after it. @jenshnielsen made it private and the issue just vanished.

(By the way, is it normal that test_boxplot_rc_parameters has a tolerance of 1?)