This is green on Travis, and therefore probably good to merge.

To trim this down a bit, the text changes seem consistent with those onmaster. There are a few different images:

git di --shortstat backport-lower-tolerance-2.0..master -- lib/matplotlib/tests/baseline_images/ lib/mpl_toolkits/tests/baseline_images/ 17 files changed, 5853 insertions(+), 5742 deletions(-)

which are mostly non-existent on 2.0, but there are a few more other changes.

lib/matplotlib/tests/baseline_images/test_artist/default_edges.png andlib/matplotlib/tests/baseline_images/test_transforms/pre_transform_data.{png,pdf,svg} are different due to altered view limits around some collections. I think that's from#5583, but I'm wondering why that's not also onmaster.

I'm not sure whylib/mpl_toolkits/tests/baseline_images/test_axes_grid1/divider_append_axes.svg is different.

I don't understand whylib/matplotlib/tests/baseline_images/test_axes/dash_offset.{pdf,png} exist only on master, butdash_offset.svg exists on this branch too. I don't think it should have been added here.

Thanks for looking at this.

I think#5583 never got fully backported to 2.0.x. I've done that, now I'll rebase this on the new 2.0.x and see how that pans out.

I don't understand why lib/matplotlib/tests/baseline_images/test_axes/dash_offset.{pdf,png} exist only on master, but dash_offset.svg exists on this branch too. I don't think it should have been added here.

I'm not sure I follow. It exists in both places, AFAICT.

I've redone this after backporting#5583 to v2.0.x and there were far fewer conflicts this time -- mainly just in tests that exist on master but not 2.0.x.

I found two other PRs that should have been backported to 2.0.x but weren't (we probably need a better workflow for that, but it is what it is for the present time). These PRs were#5501 and#5433.

It exists in both places, AFAICT.

They didn't exist when I looked ;)

I'm not sure whylib/mpl_toolkits/tests/baseline_images/test_axes_grid1/divider_append_axes.svg is different.

This file is still different. Maybe including#5651 here will make them more consistent?

These PRs were#5501 and#5433.

For some reason, GitHub is not show the backport of#5433 here even though it does say it's part of this PR if you go to your fork and look at the commit directly. Anyway...

For some reason, GitHub is not show the backport of#5433 here even though it does say it's part of this PR if you go to your fork and look at the commit directly.

#5433 was backported directly onto v2.0.x, and then this PR was based on that.

I'm not sure why lib/mpl_toolkits/tests/baseline_images/test_axes_grid1/divider_append_axes.svg is different.
This file is still different. Maybe including#5651 here will make them more consistent?

Unless I'm misunderstanding, I think this is no longer the case. Different from what? On this branch vs. upstream/master?

git diff master -- lib/mpl_toolkits/tests/baseline_images/test_axes_grid1/divider_append_axes.svg

I think this is green again.@QuLogic: Ithink I've addressed all that you found, but let me know if otherwise.

I am seeing 3 files different

 $git diff --stat matplotlib/master -- lib/mpl_toolkits/tests/baseline_images/ lib/mpl_toolkits/tests/baseline_images/test_axes_grid1/divider_append_axes.svg         |  3714 ++++++++-------- lib/mpl_toolkits/tests/baseline_images/test_axes_grid1/twin_axes_empty_and_removed.png |   Bin 37749 -> 0 bytes lib/mpl_toolkits/tests/baseline_images/test_mplot3d/mixedsubplot.svg                   | 16034 +++++++++++++++++++++++++++++++++++----------------------------------- 3 files changed, 9874 insertions(+), 9874 deletions(-)

I'll just forcibly pick those 3 files from master (since I think the differences are all non-visible anyway) and see how that goes...

Ok. I think we're finally good-to-go here.

I don't see where those other 3 files got pulled in

maybe#5621 should be backported to 2.0.x and then this backport rebased on top of it? That is the source of the missing png

I don't see where those other 3 files got pulled in

I checked them out directly from master and amended the commit it order to reduce the amount of large diff.

maybe#5621 should be backported to 2.0.x and then this backport rebased on top of it? That is the source of the missing png

Sure. Done.

Ah, sorry, I was confused

No worries -- I was confused as well. This massive PRs are hard to manage...