Register PublishDiagnosticsCapabilities

[zulus]

Simple fix for #1007

[rubenporras]

thanks

[akurtakov]

As this is prereq for Typescript update in WildWebDeveloper is there a chance for a quick release for this one?

[rubenporras]

Yes, no problem

[akurtakov]

Let me push the lsp4j and guava updates first.

[rubenporras]

Thanks for pointing that out. I had plan to wait for #1010. That addresses LSP4J and Guava, or is any other PR coming?

[akurtakov]

#1010 addresses LSP4J. The Guava one is trickier as we use it from Orbit and if we update it that would lead to other deps (bouncycastle, icu4j, objectweb-asm) from 2024-06 to be used or we can point to the maven artifact directly (like done here). Please have a say which one of the 3 options below you prefere:

• stay with old guava
• use new guava with other deps from 2024-06 Orbit
• use new guava from maven directly

[rubenporras]

To keep the target file simple, I would keep using the orbit repository, if that is possible. Maybe it is not such a big deal to just update the orbit repo?

[rubenporras]

The other option is to remove Guava, it does not look like we use for much, and we had a PR in the past where we tried to do the removal.

[akurtakov]

Here you are #1011

[zulus]

in general would be nice to check if all capabilities that are already supported by lsp4e are correctly registered in SupportedFeatures.java

[mickaelistria]

FWIW, this capability is new and wasn't required to get publishDiagnostics working, the behavior of the typescript language server isn't really backward compatible here, although the spec is (and that's why it has been the only LS failing).
But anyway, I agree with you that review capabilities often could be useful.

[zulus]

FWIW, this capability is new and wasn't required to get publishDiagnostics working, the behavior of the typescript language server isn't really backward compatible here, although the spec is (and that's why it has been the only LS failing). But anyway, I agree with you that review capabilities often could be useful.

yeah, which is funny because even if capability isn't available ts calculate diagnostics TS calculate this, but since 4.1 or 4.2 before send message they just check capability (i found this directly in source code, probably someone requested this)

I don't know since when LSP have publishDiagnosticsCapabilities (I don't see @SInCE is spec), probably most LS just send them