build_sphinx now checks for "build succeeded." on any line.

[Cadair]

This is to work around the effects of sphinx-gallery which prints
information about the gallery build out after the "build succeeded."
message.

The output from a sphinx-gallery build looks like this:

copying static files... done
copying extra files... done
dumping search index in English (code: en) ... done
dumping object inventory... done
build succeeded.
Embedding documentation hyperlinks in examples..
    processing: index.html
    processing: plot_aia_example.html
[done]
The build_sphinx travis build FAILED because sphinx issued documentation warnings (scroll up to see the warnings).

This effects astropy/astropy#4734 and sunpy/sunpy#1718.

ping @adrn @eteq

[Cadair]

The CI fails on this do not seem to be my fault?

[pllim]

Hmm, but why would coverage decrease?

[Cadair]

not a clue!!

[eteq]

Yep, the failures are unrelated (#222). The coverage change is real - if you look at the details it's one-line, because the coverage stuff is only looking at a small chunk of the actual code. I'm not worried about it, though.

@Cadair, can you confirm that you tested this on either astropy or sunpy and it worked? (Also, that it didn't produce any false positives?)

[Cadair]

I tested it on the sunpy gallery build and it worked. I have not tested it on a build with warnings.

[bsipocz]

👍 on getting this merged.

@eteq @Cadair - I've tested it on a build with warnings and it still returns nonzero exit code.

[bsipocz]

However, after having a second look, it seems that more builds are failing than in master, so would be nice to investigate and fix this: https://travis-ci.org/astropy/astropy-helpers/jobs/119195000

[eteq]

Good catch @bsipocz - it appears this is specifically a py 3.x problem... That is, the failure you're noting happens in 3.3,3.4,and 3.5, but not 2.6 or 2.7...

[eteq]

Aha, I think I figured it out. @Cadair, if you merge Cadair#1, I think that will do the trick and then this will be good to merge.

[Cadair]

ah of course!

[eteq]

Looks like that did the job - there's a new failure, but it's a red herring (internet hiccup), so I'll just go ahead and merge this.

(will add a changelog entry about this when I merge)

[bsipocz]

@eteq - while you're at it, you may also want to look at #225.

[Cadair]

REJOICE!

[Cadair]

Could this get in a bugfix release?