Test astropy-helpers v0.4.8 release against astropy

[embray]

No milestone for this unless later there is a need to release a new v0.4.x version for Astropy.

[embray]

sigh apparently the import of matplotlib causes havoc in the docs build :\

[embray]

Updated to a new version that I think might work around the docs build issue.

[embray]

sigh nope, somehow that didn't fix it. I'll probably have to wait 'til Monday to resolve this.

[embray]

Giving this yet another try, finally I took a look at the plot directive sphinxext to see what it does. Seems like just setting matplotlib.use('Agg') should get around this. I don't see any reason any package should want to use any other matplotlib backend during installation. @mdboom ?

[astrofrog]

@embray - just out of curiosity, what was the error before?

[embray]

Trying to import matplotlib (or probably more specifically matplotlibpyplot) was causing the documentation build to blow up since it was trying to use the Qt backend :(

[mdboom]

Yes -- forcibly setting the Agg backend during installation seems fine.

[embray]

This is all just to get around matplotlib/matplotlib#4165. Fortunately this is an obscure case. It sucks but I don't see any other way right now (and fortunately ISTM importing matplotlib doesn't really add much overhead).

[embray]

whew Well that seems to have done the trick. I can release this and v1.0.1 later today.

[embray]

Oops. I just realized this failed to include the ah_boostrap script which was somewhat important in this case. I will just make a new PR :(