bug#62694: 30.0.50; eglot-tests fails with recent pylsp

[Michael Albinus]

João Távora <joaotavora@gmail.com> writes:

Hi João,

>>         (:name "pylsp" :version "1.4.1")))
>
> This is not a version problem, Michael.  I have 1.7.2. running on my
> laptop and it passes all the tests.  You say
>
>> The initialization handshake is identical, except the server info
>
>> I have absolutely no idea about the language server protocol, and
>> whether both server replies are valid communication. If both are valid,
>> eglot shall support this.
>
> Then how, given the information that I've just given you (that pylsp
> 1.7.2 passes all the tests), do you plan to do to version checking??

I haven't proposed to check the version number of pylsp. I have just
reported the versions from the traces.

> Thank you very much for using your Debian to check this, but you're
> wasting your time.  Since you have "absolutely no idea about the LSP",
> can't you "trust" me on this one?  LSP coordinates capabilities using
> the handshake's capability statements.  You have witnessed an identical
> capability statement and yet different behaviour.  The servers are _not_
> perfect, and Debian's pylsp installation is subtly not behaving
> correctly.
>
>> If not, I believe eglot shall raise an error like "protocol mismatch",
>
> This doesn't exist in LSP.  It just doesn't exist.  There are no
> protocol versions.

I haven't said that this must be an LSP error. Just an Emacs error,
raised from eglot. Whether you call it "protocol mismatch" or
differently I don't care.

And you haven't answered the question, whther the server reply from the
Debian pylsp instance is in accordance with the LSP specs.

> João

Best regards, Michael.