fix(kernel): the bake skip demands CONTENT proof, not tag existence #2081
Loading…
Reference in a new issue
No description provided.
Delete branch "fix/bake-skip-content-proof"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
The existence-only skip (v0.1.679) reused a CI-mode image (-ci-1) for a release expecting -seanfarm43; the verify gate caught it on j95fw. The skip now reads the published image /.kernel-release and matches localversion+commit; errors rebuild. Guard P10 content-proof + contract M16; 15/15.
The existence-only skip (v0.1.679) was WRONG and the release verify gate caught it on its first firing (j95fw): the CI build publishes the SAME tag build/linux-kernel:${SHORT} with LOCALVERSION -ci-N baked in, so a release expecting -seanfarm43 reused the wrong artifact. Mode changes content under one tag — the tag is not a pure function of the source sha. The skip now reads the published image's /.kernel-release (the exact string verify asserts later, via the same crane-export method) and skips only when it carries THIS bake's localversion and commit. Any read error or mismatch rebuilds — the safe direction. Guard P10 now requires the content proof; contract M16 re-applies the j95fw defect and proves it goes red — 15/15.