Discussion: GHC minor version support

[alanz]

For some compiler versions, the bump makes a difference (e.g. 8.2.1 vs 8.2.2 has a different AST), but for most it does not from a hie perspective.

a) Is this a true statement?
b) should we limit the versions we support out of the box (make build), to being only the highest?

If we did (b), people could still install the lower number version, if needed.

And I guess we would still need to run CI on each of them.

[lorenzo]

What would be the benefit for us to stick with only the latest minor? I can only think of not having to run tests for those versions, but if that is off the table, then there isn't much more.

[alanz]

Well, there are plenty of complaints about how long the build process takes.

I think it might even be a good thing to have some sort of config file, with make/curses UI to present a checkbox list of versions that you want, and it persists that.

A given user probably only ever uses one or two GHC versions.

[lorenzo]

Is it still because people are running make for all versions instead of the only version they might be using locally? We can remove the make target to force them to pick a single version.

Mid-term the solution I have in mind is to offer binaries users can download, instead of having to build everything locally.

[lukel97]

I agree with @lorenzo that binaries are probably the best mid-term solution. Long-term I would like to see a separation of the GHC-version dependent modules into its own package to minimise the amount of rebuilding