floor_naming_hygiene's module annotation names a subject

[gunbai-bot]

Why

#11573 converted a prose data …: String row into a // annotation but left it at the END of the file. DESIGN §4c admits only a standalone leading block attached to a module-scope declaration, so the parse phase refuses every line of it:

src/v2/workflow/floor_naming_hygiene.dag:240:1: error: source annotation names no subject: no module item follows it.

Twelve refusals, on main. It reds the generated-artifact gate for every PR that merges main — observed on #11579's run, which is how I found it.

What

The block moves above floor_test_sidecar_suffix, the module's first declaration. The prose is unchanged.

Evidence

A gunbc compile over this entry reports no diagnostic naming this file; before the move it reported twelve, one per line. The gate itself is the executing check on this PR.

Reported, not fixed here

dag/gunbc/fleet/fleet_converge_workflow.dag fails the same gate on main with a non-exhaustive match missing ApprovalKeyringConverge and MtCollins1Boot. Different owner, different fix.

🤖 Generated with Claude Code

[briansrls]

NO LAND — already landed / close as duplicate.

Reviewed exact head 0bd6e746bf54e27b39d27be550c6856c1ae655a5.

The source repair is correct: std.source_annotation refuses an EOF annotation as UnattachedAtScopeEnd and its diagnostic directs the author to move it above the declaration it describes. This patch does exactly that, without changing the prose.

However, the identical floor_naming_hygiene.dag hunk already merged through gunbc#11780 as merge commit 62de065e16e7bde611c95438cb3b38c95e8723f4 at 2026-09-20 04:47:53Z. #11799 opened at 04:47:10Z, so the two fixes crossed in flight by 43 seconds. Current main already carries the block above floor_test_sidecar_suffix; there is no remaining source repair for this PR to land.

The exact-head checks are genuinely green — compiler, clippy, aggregate witnesses, and heal-generated-artifacts — but they validate a duplicate branch, not a change still absent from main.

Disposition: close #11799 as superseded by #11780; do not enqueue an empty/redundant landing.

Separately, the fleet_converge_mode_fleet_ssh_key_demand non-exhaustive match was already repaired on main by gunbc#11776, so that reported main breakage is also no longer outstanding.

[gunbai-bot]

Superseded by #11780, which landed the identical move 43 seconds after this opened (this PR 04:47:10Z, #11780 merged 04:47:53Z). Current main already carries the block above floor_test_sidecar_suffix, so landing this would be a redundant no-op.

The other refusal I reported in the body is also gone: #11776 repaired fleet_converge_workflow's non-exhaustive match with the ApprovalKeyringConverge and MtCollins1Boot arms.

Leaving the finding recorded rather than silent: the breakage came from #11573, where a prose row was converted to a 4c annotation and left at EOF, and the local gunbc run route accepted it because it does not execute the phase that enforces annotation grain.

— sent from proud-tern-736