'Docs for symbol' on the closing parenthesis shows last argument instead of function docs

[Houkime]

Expected:

space+k opening the function doc popup when cursor is on function call parentheses or inside them.

What actually happens:

Why important:

Prevents easy lookup of function args while filling in the call.

[the-mikedavis]

Which language server is this? Hover on parens gives the function documentation with rust-analyzer. We only show in hover what the language server gives so this should most likely be reported upstream to the language server repo.

[Houkime]

this is rust-analyzer 0.0.0 (366bd7242 2022-06-12)
https://archlinux.org/packages/community/x86_64/rust-analyzer/

[Houkime]

ok, after more investigation, this is a bit more nuanced.

a proper function call with all args in place, first parenthesis

the same call, last parenthesis (look, it shows the argument, not the function)

the empty call, first parenthesis works as needed

but the LAST parenthesis does not.

[the-mikedavis]

Does that draw_circle() line need a trailing semicolon to finish the statement? This behavior might be produced by invalid syntax.

[Houkime]

Does that draw_circle() line need a trailing semicolon to finish the statement? This behavior might be produced by invalid syntax.

Semicolon does not change anything.

These () are Rust's void. See, when the last parenthesis is inspected on the healthy call of clear_background, it shows info about the DARKGRAY color, ie, about the argument inserted into that call. If we paste something into this last call to draw_circle, see what happens:

[Houkime]

kak-lsp (written in Rust) has lsp-auto-signature-help-enable which covers filling in long function calls.
https://github.com/kak-lsp/kak-lsp
It displays the signature in a line at the bottom of the screen.

[sudormrfbin]

Signature help is being worked on in #1755.

[the-mikedavis]

The parens appear to be expected behavior from rust-analyzer: it's showing the return type of the call when hovering the trailing paren. There's an issue for signature help (plus the pr) #1331 so let's close this in favor of that issue