Should we have integration tests that use separate executables?

[hyperthunk]

This is quite awkward to set up in travis, though on my local machine using a docker image it seems to work fine. Might be an additional use case for the next major version.

[akilegaspi]

I'm a new user of distributed-process but I agree, should have at least different executables for tests such that it will also serve as a demo for new users for proof-of-concept but if maintaining it would be a drag/liability then it would be better to just create documentation for them.

[joelmccracken]

Personally I think having some integration tests/demos that actually build with the current versions of published packages would be a huge boon. Several times I've set up little demo projects myself and ended up dealing with the "I need to add how many packages to the stackge package set??" process, and it made me wonder if I was doing something wrong in setting it up. Also, examples on https://haskell-distributed.github.io/ were broken as of quite recently (and maybe still are?), although they weren't hard to fix, just some changes were made to how the network portion should be set up IIRC

[LaurentRDC]

Personally I think having some integration tests/demos that actually build with the current versions of published packages would be a huge boon. Several times I've set up little demo projects myself and ended up dealing with the "I need to add how many packages to the stackge package set??" process, and it made me wonder if I was doing something wrong in setting it up. Also, examples on https://haskell-distributed.github.io/ were broken as of quite recently (and maybe still are?), although they weren't hard to fix, just some changes were made to how the network portion should be set up IIRC

I was thinking of working on Cloud Haskell examples at AmeriHac, depending on what tracks are selected