send shutdown/exit to LSP server upon vim exit

[castedo]

• remove unnecessary wait loop after sending LSP exit
• remove old unused global variable LSPServers

[castedo]

I'm not 100% sure, but I'm pretty sure that it is not needed and does nothing more than add a 2 seconds delay (which is why this wasn't called during vim exit). The LSP request of shutdown is already synchronous so I strongly suspect this provides all the delay that is needed (and I see pylsp respond VERY quickly). After that I don't think vim really needs to wait after the LSP exit getting sent.

[castedo]

that it is not needed

Sorry, that was ambiguous. By 'it' I mean the delay wait loop code that I remove in this PR (so that the function can be called upon vim exit without issue).

[yegappan]

I'm not 100% sure, but I'm pretty sure that it is not needed and does nothing more than add a 2 seconds delay (which is why this wasn't called during vim exit). The LSP request of shutdown is already synchronous so I strongly suspect this provides all the delay that is needed (and I see pylsp respond VERY quickly). After that I don't think vim really needs to wait after the LSP exit getting sent.

This delay is needed when restarting a language server using the :LspServer restart command. The restart command stops the server, waits for the server to stop and then starts the server again. Without the delay, when starting the server again, the previous instance of the server may still be running.

[castedo]

Oh my, I should address that. I am thinking I should fix this PR so there are two distinct kinds of shutdown/exit. A "fast" one that is only used for vim exit and the existing "slow" one that is used otherwise. @yegappan What do you think?

[yegappan]

Oh my, I should address that. I am thinking I should fix this PR so there are two distinct kinds of shutdown/exit. A "fast" one that is only used for vim exit and the existing "slow" one that is used otherwise. @yegappan What do you think?

Yes. Sounds good to me.

[castedo]

OK, I've reverted StopServer with its wait/delay to be like main. This PR now has vim exit take a separate code path through a new "fast" function.

[castedo]

@yegappan I don't see any issues with this PR now. Is there anything you see that should be updated?

[castedo]

@yegappan Here's a new proposal I've coded in the PR now: a sync call to shutdown with a 125ms timeout. Seems to me a good simple trade-off. Under normal circumstances servers will respond much faster than that and vim and server are doing a proper LSP protocol dance.

Under unusual circumstances, a user won't even notice since it won't be more than an eighth of a second before the whole subprocess is terminated anyway.

[Copilot]

Pull request overview

This PR updates the Vim LSP plugin to send a fast shutdown + exit sequence to all running LSP servers during VimLeavePre, and removes an unused global server list in favor of using the existing filetype→server map.

Changes:

• Add a VimLeavePre autocmd to trigger a fast shutdown/exit of all LSP servers.
• Implement FastShutdownServer() / FastShutdownExitAllServers() using a short synchronous timeout for shutdown.
• Remove the unused global LSPServers list.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 4 comments.

File	Description
plugin/lsp.vim	Hooks Vim exit (VimLeavePre) to trigger the new fast shutdown routine.
autoload/lsp/lsp.vim	Implements fast shutdown/exit helpers and removes the unused LSPServers global.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[yegappan]

Thanks for the patch. Can you address the comments from the copilot review?

[yegappan]

I will look into adding a timeout parameter to the lsp_server.rpc() call in a later PR, so that it can be used directly.

[castedo]

Thanks for the patch. Can you address the comments from the copilot review?

Did you actually read the AI comments from the copilot review?

[yegappan]

Thanks for the patch. Can you address the comments from the copilot review?

Did you actually read the AI comments from the copilot review?

Yes. I did go through the comments added by copilot. As some of the comments are valid, I requested you to address them. Thanks for addressing the comments.