Give the process a chance to terminate correctly

[flazzarini]

Should fix the issue described in #660

[florisla]

The CI build fails because Popen.wait() does not have a timeout argument in Python 2.7.

See https://docs.python.org/2/library/subprocess.html#subprocess.Popen.wait

Would it be dangerous to simply call wait() in case of python <3 ?

[flazzarini]

Yes I didn't think of python 2 support at all, but as python 2 is end of life on 01.01.2020 will you also drop support for it in the invoke library? If this is the case this fix could applied on a python3 only branch.

[ralphie0112358]

I'd love to have this merged and released! Any updates???

[gbansaghi]

I would argue that wait() with no timeout would be the correct approach. In practice, 3 seconds should be enough, but the library shouldn't rely on such an arbitrary timeout value (what if it's run on a very old machine, what if there's really heavy I/O, etc.)

[krupan]

I'd like if we didn't hard-code the timeout to an arbitrary 3, but this is solving the problem I'm having.

[bitprophet]

Thanks for this & the additional reports!

I can't drop Python 2 just yet, and I'm also just uncomfortable with a timeout-driven solution, I'd rather attempt to identify the race condition in play & solve it with a threading event or similar; so I'm probably going to supply my own patch instead of merging this directly (tho I reserve the right to backpedal furiously if that goes nowhere 😂 ). Please follow #660 for details.