Hi folks 👋. When I'm trying to use a

docker_image

target as a runnable dependency in

adhoc_tool

or

shell_command

like so

docker_image(
    name = "image",
)

shell_command(
    name = "foo",
    command = """
    image-builder
    """,
    runnable_dependencies = [
        ":image-builder",
    ],
)

And trying to run foo using

pants export-codegen //:foo

I get

the input device is not a TTY

If I explicitly add

--interactive=false

to the docker_image then it works for the build target:

docker_image(
    name = "image",
    extra_run_args = [
        "--interactive=false",
    ],
    image_tags = ["dev"],
)

But then I'm unable to run a shell using

pants run //:image -- /bin/bash

which can be very helpful for debugging. Is it possible to add additional docker run flags when doing

pants run

? Or is it possible to not automatically add the

--interactive

flag when the docker_image is being used by another build rule? Is this something I can (automatically) do myself?

have you seen below docs for the docker subsystem? https://www.pantsbuild.org/2.21/reference/subsystems/docker#run_args i think that should cover your use case …

The arguments for the image entrypoint may be passed on the command line after a double dash (

--

), or using the

--run-args

option.

Defaults to

--interactive --tty

when stdout is connected to a terminal.

Interesting, thanks!