c8d/list: Match graphdriver dangling filter behavior#46856
c8d/list: Match graphdriver dangling filter behavior#46856vvoland wants to merge 1 commit intomoby:masterfrom
dangling filter behavior#46856Conversation
The graphdriver implementation for treats digested but untagged images (for example pulled by digest) as a dangling images for `dangling=true` filter. The `dangling=false` should not filter out these images though. Adjust the containerd image store implementation to match the old behavior. Signed-off-by: Paweł Gronowski <pawel.gronowski@docker.com>
|
I recall seeing ticket/report which mentioned it that would be addressed by this PR but I can't find it now. Anyone remembers which one it was or if I'm hallucinating? 🤔 |
|
I think we should disable this test for containerd, an image being dangling and not dangling at the same time is very weird and I don't like it |
|
Opened #46868 |
|
I think disabling for now is fine. Getting it working like the previous list would probably involve taking multiple passes through the image list to determine whether an image reference doesn't have a corresponding tagged. Simply checking whether a single reference in containerd is untagged does not match what the UI may consider to be a dangling/untagged reference. Long term either moving away from the existing "dangling" definition in the UI probably makes sense but another cleaner solution could also be returning a larger set of references and letting the cli define/implement the reference collapsing and "dangling" filter. |
The graphdriver implementation for treats digested but untagged images (for example pulled by digest) as a dangling images for
dangling=truefilter. Thedangling=falseshould not filter out these images though.Adjust the containerd image store implementation to match the old behavior.
- What I did
- How I did it
- How to verify it
TestListDanglingImagesWithDigeststest result- Description for the changelog
- A picture of a cute animal (not mandatory but encouraged)