[clangd]: what if the software is built in a docker?

[ubaldot]

I have a docker container where I installed clangd .
How to connect the LSP client to the container running the server?
Typically I specify path: "/usr/bin/clangd" when registering the LSP server. But what if the LSP server is running on a docker? Can I set e.g. path: "ssh -p 2222 root@localhost 'clangd'"? Shall I also expose ports for the communication between the LSP server and client? From the server side I think it should be enough to call e.g. clangd --query-driver=port:5000 but how to specify that on LSP client side?

[ubaldot]

I can self-answer myself.

Create a script like the following:

#!/usr/bin/env bash

set -o errexit
set -o nounset


image=my_image:local
timestamp=$(date '+%y%m%d_%H%M%S')
container_name=$(echo ${image} | sed 's/\:/\./g')

clangd_options="--log=verbose --background-index -header-insertion=never"
clangd_mappings="--path-mappings=/home/myname/proj_example=/app,/home/myname=/host_home"

# enter container
docker run --name "${container_name}_${timestamp}" -h $(hostname)-docker -i -e HOST_USER_ID=$(id -u) -e HOST_USER_GID=$(id -g) -v $HOME:/host_home -v "$PWD":/app   -e USER_ADDED_ENV="${USER_ADDED_ENV:-}" -w /app "$image" clangd "$clangd_mappings" "$clangd_options"

Needless to say that such a script can be optimized, for example you can directly mount volumes with the correct path instead of using --path-mappings, but it is good enough to show how to exploit a clangd server running in a container.

Then, you register the LSP as it follows:

    {
        name: 'any_name',
        filetype: ['c', 'cpp'],
        path: /path/to/your_script.sh,
        args: [],
    },

Few things to remark:

1. clangd must be running in the container,
2. you must have the source code both locally and in the docker along with the libraries used,
3. the value or args when registering the LSP shall be an empty list. All the options that you want to pass to clangd shall be included in the .sh script.