Another attempt at using the lsp API for some progress reporting

[michaelpj]

No description provided.

[michaelpj]

Okay, I think the tests should pass now. I had to get rid of some of the functional tests because they test the real HLS executable, so that now includes a delay on starting progress. So we can't observe it unless we can make things take long enough, and we can't make the delay shorter because we're using the real HLS (unlike the ghcide tests which use a special executable with different options). I was able to arrange for one of the tests to have a long enough delay, so that might have to do for now.

As far as I can tell, this seems to work reliably. I would also like to rewrite the other special progress reporter, not sure if I'll do that in this PR or not.

[michaelpj]

Okay, many tests rely on waiting for progress notifications and are now somewhat broken. I shall continue on it.

[michaelpj]

This will fix #381, in that we won't show progress for fast operations, which will include most formatting operations.

It should also fix #3577, or at least mitigate it, in that updates will be throttled and won't be sent for short-lived "processing" operations.

[michaelpj]

Ready for a look, perhaps @soulomoon ?

[soulomoon]

Looks good !
Just want to ask one thing, so we lose some of the notification tests since we now have 1 s delay for notification ? I wonder if we can set the shorter delay for testing and gain those test back?

[michaelpj]

Yes, so we have some special arguments for the ghcide executable that get used during testing that set the delay to 0. But for the HLS functional tests we don't call the main function of HLS we actually run the built executable, so I'm not sure how to make that nicely configurable...

[soulomoon]

Yes, so we have some special arguments for the ghcide executable that get used during testing that set the delay to 0.

So it is using

hlsLspCommand :: String
hlsLspCommand = hlsExeCommand ++ " --lsp -d -j4"

Maybe we can generalize the special arguments for the ghcide to the hls-exe ?

[michaelpj]

Yeah, or maybe I can use an environment variable or something.

[konn]

LGTM as for Splice Plugin! 👍

[michaelpj]

We already had a --test argument to HLS that turns on the testing stuff, which I had already made change the delay. So I just needed to pass that in the test invocation and it works again.

[fendor]

What a weird number. Shouldn't this be 1_000_000?

[michaelpj]

🤦