feat(docker): integrate cuda/no-cuda jobs into single job

[youtalk]

Description

Resolved #5082

As a result of the activity autowarefoundation/autoware_universe#8695, the only packages dependent on CUDA are the sensing/perception components of autoware.universe. Therefore, the container images with the -cuda suffix are actually only needed for universe-sensing-perception-devel and universe-sensing-perception.

Additionally, since there is currently only one arm64 self-hosted runner, the no-cuda/cuda jobs in the docker-build-and-push-arm64 workflow experience double the waiting time.

This PR consolidates the no-cuda/cuda jobs and changes the process so that only the CUDA dependent packages in the sensing/perception component are built in an additional stage.

Tests performed

https://github.com/autowarefoundation/autoware/actions/runs/11718481450
https://github.com/autowarefoundation/autoware/actions/runs/11718483615

Effects on system behavior

Not applicable.

Interface changes

Pre-review checklist for the PR author

The PR author must check the checkboxes below when creating the PR.

• I've confirmed the contribution guidelines.
• The PR follows the pull request guidelines.

In-review checklist for the PR reviewers

The PR reviewers must check the checkboxes below before approval.

• The PR follows the pull request guidelines.

Post-review checklist for the PR author

The PR author must check the checkboxes below before merging.

• There are no open discussions or they are tracked via tickets.

After all checkboxes are checked, anyone who has write access can merge the PR.

[youtalk]

@wenrir

Additionally, have you considered using matrix instead? In order to combine the cude and non-cuda workflows? they seem to do basically the same thing.

One of the purposes of this PR is rather to stop the matrix, because in this organization, there is currently only one arm64 self-hosted runner available, and using the matrix would require double the execution time.
https://github.com/autowarefoundation/autoware/pull/5363/files#diff-de0ada9957e6211b26d1e4d4d49c0525a9f36224a5b8c743b5ff7339130f2aabL17-L34

[wenrir]

@youtalk, thank you for the quick action and response, I understand (and agree) with the feedback.

(I'm unable to resolve my own comments, sorry)

[oguzkaganozt]

@youtalk I think in overal terms the PR solves the complicated nature of docker-build workflows and better utilizes the cache by seperating CUDA and non-CUDA builds.

But in the other hand Dockerfile would become rather complicated to track and maintain, since it is not preferrable to have multiple Dockerfiles I suggest putting some of the recurring scripts to a directory.

For example:

docker/
├── Dockerfile
├── scripts/
│   ├── install-rosdep-dependencies.sh
│   ├── build-ros-packages.sh
│   └── cleanup.sh
└── etc/
    └── ros_entrypoint.sh

I think this kind of improvement can be done in a seperate PR, and I can draft a PR for that after this PR.

And also with this PR we are not doing health-check builds with CUDA anymore as i can understand or am i missing something ? And if so would it miss some CUDA related build fails ?

[youtalk]

@oguzkaganozt I've fix a minor cache problem c009784. I hope the CI jobs will be succeeded.

https://github.com/autowarefoundation/autoware/actions/runs/11718481450
https://github.com/autowarefoundation/autoware/actions/runs/11718483615

@mitsudome-r @xmfcx docker-build (cuda) and docker-build (no-cuda) will no longer be executed, so please remove them from the required CI checks if you approved this PR. And after this PR is merged, please add a required check for docker-build.

[xmfcx]

Removed ✔️

[oguzkaganozt]

LGTM

[youtalk]

@mitsudome-r @xmfcx docker-build (cuda) and docker-build (no-cuda) will no longer be executed, so please remove them from the required CI checks if you approved this PR. And after this PR is merged, please add a required check for docker-build.

@mitsudome-r @xmfcx Please add a required check for docker-build.

[xmfcx]

@youtalk -san, docker-build is a composite action, not a workflow.

Do you mean to add docker-build-and-push as a required workflow?

But that workflow only runs on:

on:
  push:
    branches:
      - main
    tags:
  workflow_dispatch:

not every PR.

What should I do then?

[xmfcx]

Sorry, I've just realized it was part of the health-check.
I've added it to the required checks, tested in:

• feat(.github): split docker-build-and-push step to no-cuda and cuda steps #5414

[youtalk]

@xmfcx Thank you!