Description
I'm using a custom manylinux-based container image in CI.
The environment already contains:
- the manylinux userspace
- all required Python interpreters
- the complete build toolchain
When running cibuildwheel inside that container, it still requires a container engine (docker or podman) and attempts to launch a nested manylinux container.
This fails in environments where Docker is unavailable, for example:
- GitHub Actions container jobs
- some self-hosted runners
- restricted CI environments
Question
Is there currently any supported way to tell cibuildwheel:
"I am already running inside a valid manylinux container, please build directly in the current environment without spawning another container."
Something like a native/containerless Linux mode.
Motivation
This would allow projects that already maintain custom manylinux images to:
- avoid Docker-in-Docker
- avoid Podman setup
- reduce CI complexity
- reuse existing build environments directly
If this is intentionally unsupported, could you explain the main technical or policy reasons behind that decision?
Thanks!
Description
I'm using a custom manylinux-based container image in CI.
The environment already contains:
When running cibuildwheel inside that container, it still requires a container engine (
dockerorpodman) and attempts to launch a nested manylinux container.This fails in environments where Docker is unavailable, for example:
Question
Is there currently any supported way to tell cibuildwheel:
Something like a native/containerless Linux mode.
Motivation
This would allow projects that already maintain custom manylinux images to:
If this is intentionally unsupported, could you explain the main technical or policy reasons behind that decision?
Thanks!