`devcontainer exec` tries to use different container than the response of `devcontainer up` in WSL

devcontainer exec tries to use different container than the response of devcontainer up.

$ devcontainer up --workspace-folder .
[14 ms] @devcontainers/cli 0.12.1.
{"outcome":"success","containerId":"66d0453b1f8340282242c2efeff1e79af03734b45d56cfc7630b5b92b108880e","composeProjectName":"masked","remoteUser":"vscode","remoteWorkspaceFolder":"/workspace"}
$ devcontainer exec --workspace-folder . pwd
Shell server terminated (code: 1, signal: null)

Error response from daemon: Container 7a07354654d1bc329e8ba1ba2721582171fe483c541e5ee65f042c9dc6d7fe4f is not running

{"outcome":"error","message":"An error occurred running a command in the container.","description":"An error occurred running a command in the container."}

Version info: devcontainer@0.12.1 /home/sarisia/.asdf/installs/nodejs/18.5.0/.npm/lib/node_modules/@devcontainers/cli

[jbcpollak]

I had this problem when first trying this project out, with this process:

• Started VS Code and launched my dev container
• did a Git pull - at this point the running dev container was out of date but still running
• ran devcontainer up --workspace-folder ., which reported success:

❯ devcontainer up --workspace-folder .
[30 ms] @devcontainers/cli 0.22.0. Node.js v16.6.1. linux 5.10.60.1-microsoft-standard-WSL2 x64.
Running the initializeCommand from devcontainer.json...

[95 ms] Start: Run: /bin/sh -c .devcontainer/setup_container.sh

{"outcome":"success","containerId":"33db7e5053ce8679a3e55b5536a8596996899a09036eb31f51d7aa58b5d3f791","composeProjectName":"project_devcontainer","remoteUser":"1000","remoteWorkspaceFolder":"/home/project"}

• ran devcontainer exec --workspace-folder . /bin/bash and got this:

❯ devcontainer exec --workspace-folder . /bin/bash
Error: Dev container not found.
    at bailOut (/home/jpollak/.nodenv/versions/16.6.1/lib/node_modules/@devcontainers/cli/dist/spec-node/singleContainer.js:385:11)
    at doExec (/home/jpollak/.nodenv/versions/16.6.1/lib/node_modules/@devcontainers/cli/dist/spec-node/devContainersSpecCLI.js:807:43)
    at processTicksAndRejections (node:internal/process/task_queues:96:5)
    at async exec (/home/jpollak/.nodenv/versions/16.6.1/lib/node_modules/@devcontainers/cli/dist/spec-node/devContainersSpecCLI.js:741:20)
{"outcome":"error","message":"Dev container not found.","description":"An error occurred running a command in the container."}

Manually stopping the container, rebuilding it using the CLI and relaunching, then starting VS Code and separately running devcontainer exec ... seems to work.

If the container is launched first via VS Code, devcontainer exec ... cannot find the container.

The workaround I’ve found is to manually stop the devcontainer (if it was built with VSCode) then run: devcontainer up --remove-existing-container --workspace-folder . and finally: devcontainer exec --workspace-folder . /bin/bash

My goal is for my team to be able to use VSCode or the CLI/Terminal interchangeably with the same container.

using the explicit --container-id parameter on the exec command seems to work around this, although I still have trouble with #246

[chrmarti]

The extension is using the UNC path for "devcontainer.local_folder", so it can distinguish between the same folder in different WSL distributions. The CLI inside a distribution is not aware of this.