patches/driver/bes2600/*-danctnix: reconstruct from cleanups (closes #29) #33
Reference in New Issue
Block a user
Delete Branch "claude-noether/kernel-agent:noether/kernel-agent-29-per-series-reconstruct"
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?
Closes #29 — proper per-series mirrors
Replaces the broken DKMS-path mirrors from #17 with reconstructed in-tree-path series-dirs generated from
marfrit/bes2600-dkms-mobiancleanupsbranch viagit format-patch fe73571..cleanups --no-merges+ sed prefix rewrite. Retirescumulative-c5x-danctnix/fromfleet/ohm.yamlincludes (kept on disk for one cycle as bisection reference).Phase 0 inventory result
The c5x cumulative was generated from the
cleanupsbranch (29 non-merge commits) +bh-c-fossil-cleanup(1 cmt = Patch H). Base for the entire stack:fe73571183cd(the merge-base of all bes2600 branches with mobian, immediately before the Patch A merge into mobian).Reconstruction (32 patches total in cumulative)
Apply order (matches cleanups commit chronology)
Debian-metadata commit
0012(debian/copyright FSF address) is intentionally skipped — it's packaging metadata, not a kernel change.Self-consistency
The new cumulative is bigger than the c5x interim (276 079 vs 162 716) because each per-series patch carries its full
From <sha> Mon Sep 17header + commit message — vs the interim cumulative which was a single squashed commit's diff body.Verification plan (Phase 7)
Building
marfrit-packagespkgrel=6 from this manifest now. Ifbes2600.ko srcversion == 26B0003FE9F2B05DCE838C4(pkgrel=5's, currently on ohm), the per-series cumulative produces a byte-identical source tree to the c5x interim — verified correct.If srcversion differs, diff source trees + adjust until equivalence (likely missing-commit or apply-order issue).
Trailers
Every series-dir patch carries a standard
-- \n2.54.0\ngit-format-patch trailer EXCEPT the last in the cumulative (besser#18'squeue-pending-record-lock-bh-danctnix/0001-...patch). Still brittle —kernel-agent#31(ka-promote trailer normalisation) tracks the proper fix.Followups still open after this lands
-danctnixsuffix (the barepm-state-resync/,tx-sdio-dma-oob/, etc.) still have the original broken patches from #17. Out of scope here since ohm doesn't include them — but if other fleet manifests pull them, they'll still fail. Open a separate cleanup if needed.Refs: #17, #28, #29, #31.