Matrix trimming should account for new tags

[mthalman]

When a new tag is added to the manifest, and that is the only change for a given build, matrix trimming will end up trimming out a build for that image, preventing that new tag from being published. The new tag will only be published when a change occurs that would cause the image to be built, such as a Dockerfile change or base image update or disabling caching.

In a scenario like this, the trimming logic should account for newly added tags and cause the build job to run so the tag(s) can be applied.

Example build where this occurs.

I couldn't figure out the best area label to add to this issue. If you have write-permissions please help me learn by adding exactly one area label.

I couldn't figure out the best area label to add to this issue. If you have write-permissions please help me learn by adding exactly one area label.

[lbussell]

I wonder if there is any way we can have some shared infrastructure for testing these specific scenarios. We have lots of code dependent on Manifest and ImageArtifactDetails (image-info.json) at different points in the build. It would be nice if we had a set of scenarios...

• Adding images
• Adding tags
• Removing images
• Removing tags
• Cached images
• No-cache builds
• etc.

...that we could apply via tests to all applicable commands, such as EOL annotations. That way the next time we implement something like this, for example image signing, we could know right away if it does The Right Thing in all of these known edge cases.

Perhaps just a list of those edge cases is sufficient to guide writing new tests.

[lbussell]

I wonder if there is any way ...

I filed #1495 to track this.