Design 47 G: the lead's answer A-484 (one place answers the manifest of the build a module runs) #712

Merged
mesh-tools merged 1 commits from design/47-G-lead-answers-122 into main 2026-11-17 23:33:09 +00:00
@@ -1949,3 +1949,15 @@ the principle that settles it, cited as "design 47 G A-n".
entry, as A-113 says. (Live, 2026-11-18 00:16, g14: feh 0.1.8 passed its parts check and was then held "whether feh's
kept data changes form between f4cf42de… and 2130f05a… is not known … the entry of src2:f4cf42de… is not-found"; its
running build is from before G and has no entry.)
- **A-484** (joins A-480, A-480a, A-483 and A-461d into one rule; the operator, 2026-11-18 ~00:35 CET: "yes" to the
thorough fix; principles 2, 3 and 7) The node-engine answers "the manifest of the build a module runs here" in one
place, and every step that needs it (what the module provides, what it keeps on the bus and whether that changes
form, the artifact store's holder, the witness's kept and draining checks, the apply of a process) reads that place
only. It answers, in this order: the entry of the running build; else the composition this node runs of that build
(the latest kept revision whose digest the module's have names, whatever the want's build); else the module's trunk
manifest through the seat `mesh-catalog`'s verb `module`. It says which source answered, once per module and source
(principle 3), and a step that finds no answer from any source waits with words naming the three it tried. A module
counts as running here as A-482 says. The node-engine's tool `composition` reads what runs of a held module through
the same place, also when its want's build has no composition (principle 4). (From the feh proof, 2026-11-17/18: four
steps in a row held feh 0.1.8 on g14 because its running build, and about 70 others there, were placed before G and
have no entry; each was given its own fallback.)