There seems to be a memory leak in the denops server

[Warashi]

There seems to be a memory leak in the denops server

This is all that init.vim does

let g:denops_server_addr = '127.0.0.1:32123'
set runtimepath+=~/.cache/nvim/dein/repos/github.com/vim-denops/denops.vim

The shared server startup command is this

❯ pwd
/home/ubuntu/.cache/nvim/dein/repos/github.com/vim-denops/denops.vim
❯ deno run -A --no-lock . /denops/@denops-private/cli.ts
Listen denops clients on 127.0.0.1:32123

This is the version of deno

❯ deno --version
deno 1.37.1 (release, aarch64-unknown-linux-gnu)
v8 11.8.172.6
typescript 5.2.2

The result of :checkhealth denops

denops: health#denops#check

- Supported Deno version: `1.32.0`
- Detected Deno version: `1.37.1`
- OK Deno version check: passed
- Supported Neovim version: `0.8.0`
- Detected Neovim version: `0.10.0`
- OK Neovim version check: passed
- Denops status: `running`
- OK Denops status check: passed

The result of :version

:version
NVIM v0.10.0-dev-7e94074
Build type: Release
LuaJIT 2.1.1693350652
Run ":verbose version" for more info

How to check reproduction

Repeat starting and stopping nvim while watching the memory usage of the shared server with htop

[Warashi]

commit hash of denops.vim is this

❯ pwd
/home/ubuntu/.cache/nvim/dein/repos/github.com/vim-denops/denops.vim
❯ git rev-parse HEAD
68f607d93a9758d15c0d63d3268aa420ddd29edd

[Warashi]

example of RSS change is below

1. initial usage: 51804
2. start nvim: moving around 59012 - 59748
3. close nvim: 61164
4. start nvim: moving around 60884 - 62312
5. close nvim: 61762
6. start nvim: moving around 61544 - 62548
7. close nvim: 61508

P.S. using my usual configuration, example of RSS change is below

1. initial usage: 50572
2. start and close nvim: 124M
3. start and close nvim: 183M
4. start and close nvim: 215M
5. start and close nvim: 232M

[lambdalisue]

📝

I'm not entirely certain, but it appears that this could be an issue with Deno itself. I'd like to express my gratitude to @Warashi for conducting a thorough investigation. It's possible that the following issues are related.

denoland/deno#18414
denoland/deno#8327
denoland/deno#6946
denoland/deno#6116