From b37cb3985d00ce691eb6d86f2a4fe82b7294b2e1 Mon Sep 17 00:00:00 2001 From: jochen Date: Sat, 21 Nov 2026 22:26:40 +0100 Subject: [PATCH] Design 47 H: H-A-347, the Node build module's tests and the toolchain fallback --- 03-DESIGN/01-to-be/47-milestones/H-lead-answers.md | 8 ++++++++ 1 file changed, 8 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 201e9201..82b8f111 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 @@ -2151,3 +2151,11 @@ Where two preparers disagreed, the chosen one is named. the shadow's field), and the rest of that test's kind is a small fix of its own on a branch from main, reviewed and merged apart, so the operator's allowance is not held by trunk's flakiness. rules-are-weighed (a cheap split; the thorough fix starts in the same turn), fix-a-bug-met-twice. +- **H-A-347 (the Node build module's tests and the toolchain fallback, 22:26)** The three tests failed since the + desktop installed the current toolchain lock at 19:26: the lock pins no SDK since G §571 (the fetch step places + it), the tests skip the fetch and had read the SDK from a stale toolchain through the "installed last" fallback for + ten days. (1) A whole-module test supplies every input the step it skips would place (here a small SDK with the + stdio export in the workspace's node_modules), never read from what a machine happens to keep; fixed so. (2) The + "installed last" fallback of ADR 0533 §3 stands for a build, and writes one journal line naming the toolchain it + took and the lock it stood in for, so the next such stand-in is readable (H2, the Node build module). P3, + P-failure-loud, nothing-moved-in-five-minutes (make it visible).