intero doesn't want to die

[parsonsmatt]

Intero's process doesn't seem to respond well to threats, or, uh, killing. I'm noticing that there are some stray intero processes on my system that vim isn't shutting down adequately.

Open a fairly large project (eg 30-40 modules), let Intero load some of it, then quit vim. The intero process will still be running, and will also gobble up huge amounts of memory until manually killed.

I suppose we could get the PID of the Intero process and spam it with a bunch of kill signals until it dies on exit.

[jez]

That's odd for a number of reasons:

• I'd assume that Neovim spawned processes remain sub processes, and get killed when Neovim exits
• The jobstop() call sends SIGTERM, waits 2 seconds, then sends SIGKILL. Maybe jobstop() isn't getting called at the right times?

[parsonsmatt]

🤷‍♂️ All I know is that I've sometimes got rogue intero processes eating about 1.5GB of RAM. I tried putting more of the VimLeave jobstop() stuff in there but that didn't help.

[miladz68]

I have experienced this as well, my intero process did not die and after a few hours I noticed it took 14 GB of ram and my PC was unusable. This bug is the only issue I have had with this awesome plugin. If you manage to fix this it will be awesome.

if it helps, I am running manjaro linux 17.0.5

[ivanbrennan]

I noticed a similar issue when working in dockerized projects. The intero process is run in a container, and when I quit vim, the container is left running.

I was able to repro with a minimal project setup.

stack new foo
cd foo

cat <<EOF >>stack.yaml

docker:
  enable: true
  repo: fpco/stack-build
EOF

cat <<EOF >Dockerfile
FROM haskell:7.10
ARG local_install_root
ADD ${local_install_root}/bin/foo-exe /bin/foo-exe
ENTRYPOINT exec /bin/foo-exe
EOF

stack docker pull

cat <<EOF >~/minimalvimrc
set nocompatible
call plug#begin('~/.vim/plugged')
Plug 'parsonsmatt/intero-neovim', { 'for': 'haskell' }
call plug#end()
syntax enable
EOF

nvim -u ~/minimalvimrc src/Lib.hs

From outside vim, docker ps lists the running container. :quit vim, and a subsequent docker ps shows it still running.

Manually calling intero#process#kill() does stop and remove the container, but the VimLeavePre autocmd that calls that function isn't having the same effect. I tried using the VimLeave and QuitPre autocmd events as well, but it made no difference.

I'm running NeoVim 0.2.0, Ubuntu 17.04, Docker 17.09.0-ce, and stack 1.5.1, in case that's relevant.

[parsonsmatt]

This might be a problem with intero's handling of exit codes and it might be a neovim problem. It seems like the plugin is doing things right.

Todo: Make a reproduction that demonstrates the bad behavior and report upstream.