From c6dfd740af57149558d201744b06bc1c6791fef6 Mon Sep 17 00:00:00 2001 From: jochen Date: Fri, 20 Nov 2026 17:26:59 +0100 Subject: [PATCH] Design 47 H: H-A-162, a run's build step ask names no source --- 03-DESIGN/01-to-be/47-milestones/H-lead-answers.md | 7 +++++++ 1 file changed, 7 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 ee5e0739..dc0f7883 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 @@ -822,3 +822,10 @@ Where two preparers disagreed, the chosen one is named. released version is not pinned at all; its importer's build then waits for that package's package-published by its kind-1 edge (H ยง746, H-A-157) and fails loud past the builder's bound (H-A-155), never building against a commit. P3, P7. +- **H-A-162 (a run's build step ask names no source, 17:26)** Met in the review of H-A-155: the runner copies + its pipeline's source onto every build step ask, a run on request included (H-A-85 gives a run the branch's newest + source), while the seat mesh-builds' contract says the ask's source is absent for a run on request; a run's + candidate fingerprints are never in the builder's picture, so the run would wait its whole bound and fail. The + contract stands (P7, the existing form): the runner leaves `source` off a run's build step asks, with the test that + a run's ask names none; the builder waits only for an ask with a source, as H-A-155 says, and its wait also wakes + when its view takes a new revision in, not only when an entry changes. P7, P3.