Travis was experiencing absurdly lengthy elm-make times for the trivial Elm code in the acceptance test, in the order of twenty minutes. Turns out this is caused to Elm's compiler trying to parallelize over more cores than is prudent, due to idiosyncrasies in Travis's shared environment.
We have to work around this, since Evan only merges and releases bugfixes every few years (if you're lucky). We do so by wrapping elm-make in a wrapper that lies about the number of available cpus, hardcoding it to 2. As a result, CI times drop from ~30m to ~5m.
Travis was experiencing absurdly lengthy
elm-maketimes for the trivial Elm code in the acceptance test, in the order of twenty minutes. Turns out this is caused to Elm's compiler trying to parallelize over more cores than is prudent, due to idiosyncrasies in Travis's shared environment.We have to work around this, since Evan only merges and releases bugfixes every few years (if you're lucky). We do so by wrapping elm-make in a wrapper that lies about the number of available cpus, hardcoding it to 2. As a result, CI times drop from ~30m to ~5m.
Solution copied from: https://github.com/elm-lang/elm-compiler/issues/1473#issuecomment-245704142