automated regression testing with expect about survey HOT 6 CLOSED

Hmm so i'm not a huge fan of adding such an external dependency, especially since it adds new constraints on the dev flow. Thanks to survey, I try do just over half of my work on a windows machine so I dont forget about "the other side".

I'm curious if we can pull off something similar by providing a ReadWriter to the Stdin of the readline instance and terminal package.

from survey.

Although if this existed on the side of the scripts, so I could still run them by hand and then we just have Travis run the regression tests in its guaranteed Linux environment. That would definitely solve the problem in the short run (assuming the ReadWrite is non-trivial, which it feels like it would be)

from survey.

As far as effort goes, I think it would take ~5min to create a regression suite with autoexpect which could be run on travis (which now also supports OSX as well as linux). So to me that has pretty high value with little effort. For windows that is a port maintained by ActiveState: http://docs.activestate.com/activetcl/8.4/expect4win/ which might help with quasi manual testing. I could take a stab at creating these tests and hooking them up to travis, and if you decide you want to go in a different direction we can purge these scripts.

from survey.

Alright that sounds like a good plan. looking forward to seeing what you come up with

from survey.

Blorgh, I played around with it a bit more and the expect scripts that are generated do not fail cleanly when the output does not match, so this probably wont work as a regression after all. Off to try something less easy :)

from survey.

I'm closing this since the autoplay PR was finally merged

from survey.