This one really makes clear how the vertical alignment is wrong for text with superscripts as I've been saying on gitter.

This is probably the same root cause as #6847

@mdboom: Your fix for the x-axis aligment does indeed fix the issue here but also changes the aligmnent of labels in many other test images, and some of them look worse. Maybe we should review that in a different PR?

I think regular text is now one pixel too low (in addition to exponents being too low already):

@zblz This needs a rebase

The alignment fix by @mdboom introduces changes in many many labels on the test images. I'm not sure about many of them and think this should be discussed in a separate PR. I propose to split the alignment fix into its own PR so we can discuss it independently.

This is a reasonable proposal provide an option that some will like, but I am not comfortable with the extra complexity it adds to handle the positioning. It sounds like a better way to to do that is via positioning labels by their baseline, as suggested somewhere that I can't find right now. I'm open to arguments either way.
I'm close to changing the milestone to 2.1, since this is really a new feature rather than a style change. As an intermediate step, I will use a label to make it clear this is not a blocker.

See also #2015, #4115. It would certainly be nice to get all of this resolved clearly and cleanly.

This needs a rebase and a review, so I don't think it's going into 2.0.

This looks like a neat feature! @zblz could you rebase it on to the current master branch?

@dstansby - This is now rebased on master with tests updated for py.test.

Note that I have also removed the baseline fixing commit added by @mdboom. As mentioned above, it introduces a lot of changes in baseline images that I think should be discussed in a separate PR or issue such as #4115.

Closing in favor of #8322