image/cache: fix isValidParent logic

[runcom]

Fix #31186

Signed-off-by: Antonio Murdaca runcom@redhat.com

[runcom]

I'll add a test asap, in the middle of something else now.

[vdemeester]

/cc @tonistiigi @cpuguy83

[tonistiigi]

LGTM, but needs a test

[runcom]

Yes, as noted above, I'll need to add one.

[aaronlehmann]

I think isValidParent should be renamed, since it's no longer checking for a parent/child relationship.

[runcom]

I think isValidParent should be renamed, since it's no longer checking for a parent/child relationship.

I'll add another commit to change that or follow up (do you have any suggestion on the new name?)
@tonistiigi added a test

[aaronlehmann]

I'll add another commit to change that or follow up (do you have any suggestion on the new name?)

Actually, I might have misunderstood the fix.

I see that isValidParent still checks if len(parent.History) >= len(img.History) { (not >).

So it seems like it still is checking for a parent, but you fixed a problem that happens if the parent has the same number of filesystem layers. If this is the case, the name doesn't need to change.

LGTM

[LK4D4]

LGTM