Don't use the Module(...) constructor to create a module

[asinghvi17]

On the advice of the compiler team, we should step away from using the Module(name) constructor and instead eval in a module via an Expr.

This simply changes the way that happens. Fortunately we have the module GC code already so that should flag us if something is going wrong.

This currently evals the module to Main, but there's no reason it couldn't eval to somewhere else.

[mortenpi]

This currently evals the module to Main, but there's no reason it couldn't eval to somewhere else.

It looks like Module(m) puts it into Main too, right? There's quite a bit of potentially fragile code that handles the module name, so I think it's best we keep in it Main.

[fingolfin]

This sounds like a good idea to me, but the test failures seem concerning :-(

[asinghvi17]

Yeah, not sure why the tests failed, but they magically turned green on the last commit. Might have been an intermittent thing, we'll see now I suppose

[fingolfin]

It turned green because you left no code changes in :-). I.e. this PR now only modifies the CHANGELOG (resulting in a merge conflict...)

[asinghvi17]

Ah no! I thought that would revert the commit, not the file 😅

[asinghvi17]

Hmm, it looks like wrapping here somehow changed printing for errors and vectors? Maybe it's the module declaration bringing in Base?

A lot of the errors seem fine to me, some are concerning though.

[fingolfin]

Your changelog changes seem wrong, duplicating some existing entries