From 3e177edf69ac1338a387a847cd8927cf6a465404 Mon Sep 17 00:00:00 2001 From: jochen Date: Fri, 20 Nov 2026 16:23:37 +0100 Subject: [PATCH] Design 47 H: H-A-147, validate's base revision is pinned with the take-in --- 03-DESIGN/01-to-be/47-milestones/H-lead-answers.md | 9 +++++++++ 1 file changed, 9 insertions(+) diff --git a/03-DESIGN/01-to-be/47-milestones/H-lead-answers.md b/03-DESIGN/01-to-be/47-milestones/H-lead-answers.md index 547d8dc2..d3fb911a 100644 --- a/03-DESIGN/01-to-be/47-milestones/H-lead-answers.md +++ b/03-DESIGN/01-to-be/47-milestones/H-lead-answers.md @@ -712,3 +712,12 @@ Where two preparers disagreed, the chosen one is named. failed is forgotten: the engine asks `versions` again at the next want, the next `compose` by hand, and after a bound while the module is held, and says so in its journal; a resolution whose fetch succeeded is kept as today. At this clean state a restart of the desktop's node-engine stands in once. P3. +- **H-A-147 (the base of validate's "before" case is pinned with the take-in, 16:26)** Met live at pipelines 13 and + 15: the pin judge (§749, the removal of G-15 §579) removed the branch's previous revision in the very write that took + the next one in, because its own pipeline had ended and nothing else pinned it; validate (H-A-108, judged against + the branch's previous revision, read as the number before the candidate's) then could not run and the composition + was not asked. The indexer pins, in the take-in's write, the branch's previous revision (the newest it kept before + that write) for the same pipeline record as the take-in's own, and frees both with that record at the pipeline's + end; so validate's base stays as long as the pipeline that needs it. Not chosen: validate taking the newest kept + revision below its own, which would count every module moved since as altered and refuse problems the change did + not cause (ADR 0578 §2), a silent guess. P7, P12, P3.