diff --git a/03-DESIGN/01-to-be/47-milestones/H2-lead-answers.md b/03-DESIGN/01-to-be/47-milestones/H2-lead-answers.md index f4a1b36c..49d513e4 100644 --- a/03-DESIGN/01-to-be/47-milestones/H2-lead-answers.md +++ b/03-DESIGN/01-to-be/47-milestones/H2-lead-answers.md @@ -951,3 +951,9 @@ The lead's one-line answers to the questions H2's develop and review raise (impl goroutine; on a loaded build node it reads too early (main.251). A test waits, bounded by the level bound, on the signal that the work it reads after has run (here: the keeping says it took the need), never on an order it assumes. Every test of that file that reads a goroutine's result is held to the same rule in the same commit. P-evidence. +- **H-A-477 (a setting's word is never a bundle's own, 16:45)** When the node-engine rebuilds a running module's bundles + from the tool runner's environment, it leaves out every word the mesh itself said (the setting words `MESH_CONFIG_*` + as it leaves out the owned paths, the secret places and the fingerprint), so ยง781's check compares the manifest's + words with the setting's composed value and never a value the engine composed before. The config seat's resolved + value of a key is composed again when its declaration (its default) changes; until that is built, the engine's + composed value wins over a stale resolve, said in the journal. P-failure-loud, P3.