bug#66144: 29.1; eglot-shutdown request params violate JSONRPC

[Javier Olaechea]

> Can you test gopls shutdowns
with your patch Javier?

I have tested locally with gopls v0.11.0 and M-x eglot-shutdown works successfully afaict. Both w/o and w/ the patch applied.

However I think it might be better to modify jsonrpc-request so that when the param argument is nil the JSON serialization omits the param key. I'm thinking of using the `,@(when param (list param)) idiom. jsonrpc has tests. I'd like to give it a go this weekend to see if I can come up with an acceptable way to have jsonrpc.el conform with the standard w/o changing the API. What do you think?

On Sat, Oct 14, 2023 at 1:30 PM João Távora <joaotavora@gmail.com> wrote:

On Sat, Oct 14, 2023 at 7:02 PM João Távora <joaotavora@gmail.com> wrote:
>
> On Sat, Oct 14, 2023 at 6:19 PM Javier Olaechea <pirata@gmail.com> wrote:
> >
> > > If you can find some Eglot use case that is actually hurt by this, we can reopen.
> >
> > The OP has an Eglot use case that is hurt by this, they cannot shutdown the ocamllsp server. Because when they call M-x eglot-shutdown, eglot sends an invalid request. The server dies instead of returning a response and then eglot restarts the server.
>
> But what OP?  I don't see this message, can you point to it?

OK, I see the message now.  In that case that changes things.
One of the misbehaving servers has to correct itself.  I'd say it
should be gopls.

Unless that has already happened.  Can you test gopls shutdowns
with your patch Javier?

João

--

"I object to doing things that computers can do." — Olin Shivers