[Bug] Failing Additional checks since Dockerfile and docker/ubuntu/Dockerfile are different

[echoix]

Describe the bug
A clear and concise description of what the bug is.

Since #3170 modified the docker/ubuntu/Dockerfile without keeping in sync the main Dockerfile, one of the additional checks fails. It seems to have been placed exactly for this reason, to prevent diverging the two files. The PR was merged with that check indicating failure before merge.

Now every PR fails this check.

This was merged with broken Additional checks check which is actually catching an issue relevant to this PR namely enforcing intentional code duplicity.

Originally posted by @wenzeslaus in #3170 (comment)

To Reproduce
Steps to reproduce the behavior:

1. Open any new PR based on the main branch

Expected behavior
A clear and concise description of what you expected to happen.

Keep the two files in sync.

Screenshots
If applicable, add screenshots to help explain your problem.

System description (please complete the following information):

• Operating System: [e.g. Windows, Linux, ..., incl. version]
• GRASS GIS version [e.g. 7.8.4]

Additional context
Add any other context about the problem here.

[neteler]

The "diff" test is here:

grass/.github/workflows/additional_checks.yml

Line 45 in 0bc5039

- name: Check that files with the same content are the same

How about making this a "required" test?

[echoix]

The "diff" test is here:

grass/.github/workflows/additional_checks.yml

Line 45 in 0bc5039

- name: Check that files with the same content are the same

Yep

How about making this a "required" test?

It could, but does all the additional checks need to pass? If not, the easiest might be to add that single step to the top of a required check, like the ones that build the Docker images. Since when this step could fail, it would be when changing Dockerfiles, and when we change Dockerfiles, we check that this action passes. Instead of a generic Python-related one (even if it isn't necessarily).

[echoix]

But in the meantime, this trivial PR (the changes were already longly discussed before merging the #3170) could be merged to not block any other PRs with that failing check. And that check, if ignored, prevents to see real failures that could run under it.

Edit:

Sorry I thought this was the PR, but it is the issue. See #3205 that was ready.