From 775eefcfed401dda966f5c55d8c1c93ab5d39a9e Mon Sep 17 00:00:00 2001 From: jochen Date: Sat, 21 Nov 2026 21:15:50 +0100 Subject: [PATCH] Design 47 H: H-A-326, a mover waiting on a publish that already happened is placed at once --- 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 8ed3a76d..f80017c4 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 @@ -1958,3 +1958,12 @@ Where two preparers disagreed, the chosen one is named. allow's answer form gains an optional field outward on each kept allowance, the outward link it opens, set when a named interface is one. Both in the group's contracts and networking parts. P7 (one classifier of links), P-failure- loud. +- **H-A-326 (a mover waiting on a publish that already happened is placed at once, 21:15)** Met the third time tonight + by a restart: the catalog repository's pipeline had every test ended by 20:53 and stood twenty minutes, two movers + "waiting for mesh-grants:busgrants / grantsstate before their build" although those packages were published at + 20:00, before the wait was written, so no publish event ever came for them; the runner's start pass placed them. + Fixed in the runner, on the open branch: (1) when a mover's wait on a package is written, the runner reads the + package's published state at once and places the mover when it is already there (a wait is a read first, an event + second); (2) the stalled rule of H-A-311 (3) counts a mover waiting on a package as nothing open, so such a pipeline + is named unhealthy after five minutes; (3) the journal says a wait resolved from a read apart from one resolved by + an event. P-NATS-native (read the state, then hear the change), P-failure-loud, H-A-311.