Closed catalin-hritcu closed 5 years ago
Is the --eager_inference
flag hard to maintain?
It was useful at debugging in the past, so why not just keep it.
It will be yet another configuration of the type-checker to test. I don't think our regression suite uses it at all. So, I worry that counting on it going forward will lead to unpredictability.
If you really want to keep it, I don't mind. Let's document it as "experimental" or "brittle" or something like that.
According to Nik's comment on https://github.com/FStarLang/FStar/issues/424#issuecomment-150864813 --eager_inference flag should be disabled. "It's extremely non-standard and was there to support a transition from the old inference algorithm to the current one."