Retire the gunbc#11193 anchor admission row, consumed by its own landing

[gunbai-bot]

Deletes one row and the doc paragraphs that described it. Nothing else.

Why now

#11347 landed the three bare-name import fixes at a745dac28d5. The admission row it carried admitted a namespace delta between base and head — and with the transition now present at the base, there is no delta left for it to admit. A consumed row refuses roster edits and landing (base equals head), so main's own push fails namespace-wave-admission until this lands.

This is the repayment half of the debt #11347's body declared, and it was always mine. It is paid adjacent to the borrowing rather than deferred, which is the entire reason the fix was extracted from #11193 in the first place: coupled to a blocked PR, the row would have sat in the roster across an indefinite wait and then blocked every lane's landing at an unscheduled moment.

The roster's resting state is empty. It is not a place rows live; it is a place rows pass through.

What else changed, and why it is not scope creep

The doc block above the array described the row in detail — "ONE ROW, gunbc#11193: …". Deleting the row and leaving that prose would leave a doc block asserting a row that no longer exists, which is worse than either state alone. Those paragraphs go with it.

Untouched: the general commentary on consumed rows, and the #11182 relocation distinction (a consumed row goes because its transition is present at the base, not because the array was being emptied).

Verification

• cargo check -p v1-compiler --lib — clean
• cargo fmt --all --check — clean (pre-commit hook)
• The array returns to &[], its documented resting state

🤖 Generated with Claude Code

https://claude.ai/code/session_01D2AF1pri3WJkug5AtCYLGF

[gunbai-bot]

Fixed in 329b2f2a. Review 66052 is correct and the finding is against me in the sharpest possible way: I deleted the doc paragraphs describing my own row and left the neighbouring ones asserting a population the declaration no longer carries.

Both claims are false against &[]:

• :1774 — "What remains below is the Record the contains dispatch defect, and repudiate an approval that was never granted #11182 relocation set alone"
• :1803-1809 — "these seven still admit a real delta and deleting them would refuse a live transition"

And the review's sharpest point is that this is the shape the same block condemns two lines up: "prose naming ROW ONE and ROW TWO when neither row exists is worse than no prose, because it reads as coverage." I quoted that doctrine in my own commit message while violating it one paragraph below.

On the "already stale on main" point — agreed, and it does not excuse leaving it. The seven-row claim was stale before this PR, since main's array held one row. But this file's own doctrine is that a consumed row's deletion comes due on the roster's own next touch, and this is that touch. Inheriting the stale prose to the next lane is exactly what the #11156 paragraph says not to do.

Kept deliberately: the #11156 narrative and the base-presence distinction it draws — a consumed row goes because its transition is present at the base, not because the array was being emptied. That reasoning is still true and is the reason the block exists. What replaced the stale paragraph states the roster is empty, notes the #11182 claim was already stale on main, and says why it is repaired here.

Also confirming the two checks the review made that I had not: run_wave_admission_between already carries an is_empty() → NoSubject arm, so an empty roster is a modeled state rather than an untested one; and git grep 11193 shows no mirror row left dangling in dag/gunbc/namespace/namespace_wave_admission.dag.

cargo check -p v1-compiler --lib clean; cargo fmt --all --check clean via the pre-commit and pre-push hooks.

— sent from sunny-carp-329