ADR 0262: note how a left-out module's data is backed up, as the controller now does it
mesh/merge-gate pass: the change touches no module of the mesh's graph
mesh/repo-check pass: its merge-check.sh passed
mesh/delivery delivered

The record said its data is still copied; the shipped follow-up backs it up as this controller reads
it and lists an unplaceable line rather than failing the machine.
This commit is contained in:
jochen
2026-10-08 18:38:09 +02:00
parent 79bd9911e6
commit f30b2a3ae0
@@ -133,6 +133,16 @@ refused for it, and keeps its backup lines, so its data is still copied while it
condition then says exactly what stopped. Until that follow-up is live, the condition and this record
say what stops as above.
> **Progressive insight — 2026-10-08.** The follow-up shipped (mesh-controller pull request 162) and
> two facts above are sharper than written. Where the paragraph says "its data is still copied", read:
> its data is **still backed up as this controller reads it, which may not be what its newer manifest
> asks**, because a key the controller does not know may be about that very data. And a left-out
> module's backup lines are **best effort**: a line that cannot be placed (an access nobody placed, a
> directory or machine fact with no value, a placement setting that does not read, in which case the
> definition's own path is never used instead) is listed among what the plan could not place, and is
> never an error that fails the machine's declaration. The decision stands. *Checked by*
> `TestALeftOutModulesDataIsStillBackedUp` and `TestALeftOutModulesUnplaceableBackupLineCostsOnlyThatLine`.
## Consequences
- The notifier's font size and width become preferences with today's values as defaults, and the