CI failure due to backwards-incompatible aeson change

[gelisam]

I believe the CI failures in #206 are not due to changes in my PR, but rather, to a backwards-incompatible aeson change (presumably the much publicized switch away from HashMap, due to the discovery of a collision attack).

I can reproduce this on master by running cabal update before cabal build, so this probably requires both publishing revisions with new upper bounds and updating HEAD to be compatible with aeson-2.x. And possibly some CPP magic to continue to be compatible with aeson-1.x.

[cdepillabout]

Thanks for bringing this up.

I published a revision for termonad-4.2.0.0 (the current latest version), that makes sure to put an upper bound on aeson: https://hackage.haskell.org/package/termonad-4.2.0.0/revisions/