Turn HLS-wrapper into an LSP Server #2960
Conversation
| @@ -30,7 +31,8 @@ data ReactorMessage | |||
| | ReactorRequest SomeLspId (IO ()) (ResponseError -> IO ()) | |||
|
|
|||
| type ReactorChan = Chan ReactorMessage | |||
| type ServerM c = ReaderT (ReactorChan, IdeState) (LspM c) | |||
| newtype ServerM c a = ServerM { unServerM :: ReaderT (ReactorChan, IdeState) (LspM c) a } | |||
There was a problem hiding this comment.
I couldn't get the call sites of runLanguageServer to compile without this refactoring: the compiler seemed to infer the wrong kind for implicit type variable a when using a type alias for ServerM.
smatting
force-pushed
the
hls-wrapper-lsp-server
branch
from
d654663
to
5ad4ef4
Compare
|
Is it still possible to use ghcide/hls without the wrapper? |
|
@wz1000 of course, this is just in the case that the |
|
I don't quite understand how LSP state (eg config, VFS state) updates which could potentially be intercepted by the wrapper are communicated to the main HLS process. |
|
HLS-wrapper only activates the LSP mode if there was a breaking start-up issue. For example, when we can't figure out the GHC version, consequentially can't start the main HLS process. The only way forward is still to restart HLS and the wrapper, but there is a UI pop-up that tells you this instead of silently not starting, in combination with a "Restart" button that may be displayed to the user. |
|
I don't understand what test is failing and why. There were some merge conflicts in e7dee82 but it was only code blocks that have been moved by this PR. |
* Make wrapper a LSP on failure * Fix incorrect imports * revert import block for smaller diff * add missing imports * Fix: callProcess on win32 machines not called * import container only on win32 * add missing liftIO Co-authored-by: Pepe Iborra <pepeiborra@gmail.com>
Introduces better error handling to HLS-wrapper to show LSP clients
dedicated error messages.
This should help users understand why their Language Server isn't
starting.
Closes #2589
Note: This is continuation of @fendor 's PR #2591 developed with his help at Zurihac with notable additions:
runLanuageServerhas been refactored to act as an entrypoint for both HLS LSP and the HLS-wrapper LSP. This shoud be true refactoring for the HLS LSP call siteI've manually tested the happy path (wrapper calls hls) and a error case in emacs, where restarting worked.