[cider#1052] Don't debug synchronous code

[Malabarba]

Relies on clojure-emacs/cider#1070

[bbatsov]

I wonder if this will always behave as we expect it to. Does this code always run in the context of interruptible-eval (didn't go through the implementation in much detail)? @cichli does it look legit to you?

[Malabarba]

Perhaps I should go with a more verbose name for the key, like :cider-sync?

[Malabarba]

Ah, I think I get your question now. I suppose there's nothing preventing other middleware from evaling code as well.
In fact, I think pprint might be one.

[cichli]

It seems breakpoint is only ever called within the context of an nREPL eval, so *msg* should always be bound here. If we want to call this function from other middleware handlers we could just bind *msg* to the message being handled, assuming we include the :sync key for all requests (i.e. not just eval requests).

Not been following the issue too closely, so forgive my ignorance, but shouldn't this really be fixed client-side (i.e. properly handle need-input during synchronous requests) so we don't need this kind of hack?

[Malabarba]

Not been following the issue too closely, so forgive my ignorance, but shouldn't this really be fixed client-side (i.e. properly handle need-input during synchronous requests) so we don't need this kind of hack?

Yes, this should be fixed client-side, but I'm under the impression that even with the handle-input fix in sync-request the breakpoint will still bug out during a sync eval (because its workings are more complicated than a simple need-input event). Still, I haven't had a chance to test that conjecture yet (busy busy busy :-P).