EmittedBuildNotDiscriminating carries the faulted run's stderr, so an unattributable red is located

[gunbai-bot]

The unattributed mutation-probe refusal now carries cargo's own output.

EmittedBuildNotDiscriminating named its class and dropped the faulted run's stderr, which run_cargo had already captured as stderr_tail. An unrelated error, a lock, a full disk and an unreadable rendering all printed the same sentence, so the cause was unreachable from every CI log. That is how the coloured-header defect #12091 repaired stayed unlocated until the tail was read (run 35811281201). DESIGN §5 asks for typed AND located.

Change

The refusal is built by one function, unattributed_fault_refusal, which establish_discriminating_red calls and which interpolates cargo_verdict_stderr_tail. No other behaviour changes.

Evidence (remote, executed)

• Green on this branch: an_unattributed_fault_refusal_carries_the_faulted_runs_stderr passes. A faulted Completed { status: 101 } run whose stderr carries a known marker surfaces that marker in the refusal.
• Red on main's refusal text: the same test with only the interpolation removed fails (panicked ... the refusal must carry the faulted run's stderr).

Supersedes the one piece of the closed #12105 that #12091 did not carry.

🤖 Generated with Claude Code

[gunbai-bot]

Review 70498 addressed at a380fb8:

• Doc-block misattachment (both sites): fixed. unattributed_fault_refusal and an_unattributed_fault_refusal_carries_the_faulted_runs_stderr now sit above the existing /// blocks. "THE DISCRIMINATING RED, ESTABLISHED BY MUTATION…" documents establish_discriminating_red again, and "THE SPAWN OWNS EVERY CHANNEL…" documents the probe-spawn test again. The change is a move only; no text changed.
• Seed-growth receipt (advisory): no row added, deliberately. This modifies an existing hand-authored host item's refusal text and extracts one private helper from its body. It adds no new host capability or route, so there is no new seed surface for a row to count. gunbc.emitted_closure_compile_seed_growth already carries this module, and its trigger is unaffected. As you note, the PURPOSE test is met: the change serves the v2 self-host instrument's own refusal.

— sent from smart-tern-891