Assume backend/optional package checks fail after defined timeout

[JDWarner]

Over in scikit-image we're having some trouble with our Travis build, and it appears to be due to the matplotlib optional backend package check for Qt4 entering a subprocess and never returning under the Travis system Python (3.2.3). This results in an eventual Travis timeout.

See the bottom of this build log: https://travis-ci.org/JDWarner/scikit-image/jobs/34035472

The reason this happens is unclear and it would be ideal if this worked (as PyQt4 is present in that 3.2 build), but the real problem is that this can completely tank a CI build or install. Ergo, we recommend a timeout (perhaps ~15 seconds) for checking [optional] build packages.

[jenshnielsen]

I discussed this with @JDWarner at the euroscipy sprint.I have seen a similar issue with a broken pygobject hanging the setup script

[jenshnielsen]

#3443 is a duplicate of this

[jluttine]

I've experienced this problem too with my project. Just in case it helps, I forked matplotlib and added Python 3.2 to Travis tests. Here is the failing output:
https://api.travis-ci.org/jobs/34094194/log.txt?deansi=true
If this was just repeating the same info, then sorry for spamming.

[jenshnielsen]

@jluttine thanks for sharing, Looks a bit different from the scikit image one so it is useful.

[tacaswell]

@jluttine As a side note, we intentionally dropped testing 3.2 and do not claim to support it (see #3192 )

[jenshnielsen]

While @tacaswell is right I would like to try to get to the bottom of this. I don't like that the setup script can hang like this and suspect that this might have more to do with the specific pyqt installation than the python version

[tacaswell]

Not willing to block 1.4.1 on this issue, re-milestoning to 1.4.x.

[jenshnielsen]

I believe this was fixed by #3741 please reopen if that is not the case