[Fix #810] don't allow requests before client initialization has completed

[vspinu]

Cider should not be useful before the initialization has completed. So, don't use async requests during client initialization.

[bbatsov]

If we're going down this path it makes sense to do the initial describe session sync as well.

[bbatsov]

Forget my comment, I just saw you've actually done this already.

[bbatsov]

The code looks good, but you'll have to rebase.

[vspinu]

Ok, have done that. Thanks.

[bbatsov]

Btw, we should probably invoke those sync requests without a timeout, or we might run into a problem similar to the slow version requests that were interrupting the startup in the past. Waiting indefinitely on a request, that's never going to blow up internally, is OK.

[vspinu]

Bozhidar Batsov on Fri, 19 Sep 2014 04:54:29 -0700 wrote:

Btw, we should probably invoke those sync requests without a timeout, or we
might run into a problem similar to the slow version requests that were
interrupting the startup in the past. Waiting indefinitely on a request, that's
never going to blow up internally, is OK.

I aggree. I also think we should increase the timeout on remotes. I have
experienced timeout yesterday on C-c C-n in REPL on a fully loaded
remote machine.

[bbatsov]

I have
experienced timeout yesterday on C-c C-n in REPL on a fully loaded
remote machine.

I think we should drop the timeouts completely or at least disable them by default. This will force our hand to fix quickly the middleware problems. To make this viable we should just add :status :done to the messages we send in case of an exception and send the same data we send on eval exceptions.

[vspinu]

I think we should drop the timeouts completely or at least disable them by
default.

Yes, completely agree.