Fix TC allowing bad schemas in bind, with-read, with-capability

[sirlensalot]

No description provided.

[jwiegley]

Looks great to me, I'll kick off a replay now.

[sirlensalot]

Looks great to me, I'll kick off a replay now.

This only affects static typechecker so no replay is necessary

[jwiegley]

Thanks, @sirlensalot, that should have been obvious to me.

[emilypi]

LGTM thanks @sirlensalot 👍

[imalsogreg]

Great catch! fyi I've verified that my return-types PR now passes tests when rebased on this. (although that's not surprising, because PR-1212 fixes the Pact code that didn't typecheck before).

Replay doesn't need to be involved for PR-1212, but are there any other backward compat issues we need to think about? (e.g. a customer that runs TC and accidentally relied on the previous lax TC behavior?)

[sirlensalot]

are there any other backward compat issues we need to think about? (e.g. a customer that runs TC and accidentally relied on the previous lax TC behavior?)

In general breaking repl-only compat is ok, and static TC + FV are even more so as they're supposed to be about correctness first. Finally since this goes with the runtime TC fork that certainly trumps repl compat.