Fix suptitle

[dmcdougall]

New version of #1276 but which targets v1.2.x and includes a test.

If there's no feedback I'll merge this in a day or two. I'd at least like @NelleV to check PEP8-ness, if you wouldn't mind :)

[NelleV]

PEP8 compliance: you need two blank lines here.

[NelleV]

I've underlined the two PEP8 problem. Else, the code looks good to me.

[dmcdougall]

Wooo! A legitimate test failure!

@pelson Do you have any idea why that pickle test is failing? Looks like that title in the pickle baseline image is a suptitle, but the kwargs are still being set just as they were before this change. Perhaps the addition of a self._suptitle variable affects how the figure object gets pickled.

[pelson]

@dmcdougall - I'll have a butchers tomorrow at this for you. Has the result image gained a suptitle that wasn't there before?

[dmcdougall]

@pelson Another travis downside: I can't access the failing image. I'll run the test locally and report back.

[dmcdougall]

@pelson The produced image doesn't have a suptitle. Will look into it.

[dmcdougall]

I've found the issue. Since the suptitle (a Text object) now persists in the Figure class, having remove_text=True in the image_comparison decorator for the test_pickle test removes the text when producing the resulting image.

There are a couple of ways to proceed:

1. Remove remove_text=True, resulting in a persisting suptitle but potentially causing further failures because of text issues.

2. Keep remove_text=True and remove the suptitle. There is an additional check inside test_pickle to ensure that figure labels are preserved on unpickling. Downside is a new baseline image has to be committed.

3. Don't merge this pull request.

I think 3) is overly harsh. I'd prefer to execute option 2), but if there is a consensus for 1), then I'm fine with that too.

[WeatherGod]

I am +1 on option 2.

[mdboom]

Ditto for option 2.

[dmcdougall]

Done.

[pelson]

I think this should call the .remove() method, rather than just making it a null Text object.

[dmcdougall]

Or perhaps .update, and put the line below inside an else:?

[pelson]

Either way works. 😉

[dmcdougall]

@pelson Better?

[pelson]

Yep. 👍 from me. @NelleV - are you happy with it?

[NelleV]

LGTM: 👍 for merging

[dmcdougall]

Ok. I'll clean up the commit history and merge this (py31 failure is a dud). Thanks.