Change signature of on_close hook (queue by val)

Author: g41797

design/secondary/lang/c/ctk-proposal.md

@@ -415,7 +415,7 @@ The stack's was the sharper case, and the deletion rested on two grounds.
 **The first is `R15`: `push_slot`'s only caller was ever `pool.c3:451`,
 `put_all`'s refusal path**, and 002 dropped `put_all` — §5.1's row says so, and
 the queue's `push_front_slot` went with it for the same reason. The second, from
-`3tk-port-findings-003.md`, was that no application could reach an `InnerStack`
+`3tk-port-findings-004.md`, was that no application could reach an `InnerStack`
 at all. **3tk-only, and it was two lines plus one test.**
 
 **CLARIFIED 2026-08-26.** The owner's ruling that the stack is available to a

design/secondary/lang/c3/3tk-deviations-001.md

@@ -7,6 +7,57 @@ Current state is in [3tk-status.md](3tk-status.md).
 
 ---
 
+## 2026-08-30 — 3TK-56: the close hook takes the queue by value, in code
+
+**Built [3tk-on-close-handoff-001.md](3tk-on-close-handoff-001.md), the
+2026-08-28 ruling.** `P6` closes: `on_close` receives `InnerQueue`, not
+`InnerQueue*`, and the pool holds nothing back.
+
+**`InnerQueue.take()`** — `3tk/src/queue.c3` — hands over everything the queue
+holds and empties the source, O(1), cannot fail. Both call sites in
+`pool.c3` — the straggler path (`:494`) and the main close (`:564`) — became
+`self._hooks.on_close(stragglers.take())` and
+`self._hooks.on_close(remaining.take())`.
+
+**The six implementers followed**: `negative/common.c3`,
+`negative/release_with_straggler_put.c3`, `negative/release_during_on_put.c3`,
+`negative/release_during_on_close.c3`, `test/t_concurrency.c3` (two hooks),
+`test/t_pool.c3`. Only the signature line changed in every one — a by-value
+struct parameter is an lvalue, so `&self` methods (`pop_front`, `is_empty`,
+`len`) bind to it exactly as they did through the pointer, and no body needed
+touching. The three `release_during_*` tier 1 negatives still abort in every
+build mode, unweakened.
+
+**One positive test, no negative.** `t_queue.c3:take_empties_the_source`
+proves the source is empty after `take()`. No negative is possible: a hook
+that keeps items dereferences nothing, so no build could ever notice a leak
+however the parameter is spelled — that is the ruling working as intended, not
+a gap in coverage.
+
+**Wording, not just code.** `PoolHooks.on_close`'s doc block in `pool.c3` now
+carries *the pool does not verify it and never will*, and the same sentence
+went to `ref/3tk-reference-004.md` — named directly by the charter, and edited
+in place rather than versioned. `ref/3tk-decisions-004.md` gained the `P6`
+entry under `pool.c3` and a `take()` entry under `queue.c3`; `-003` moved to
+`backup/`. `3tk-port-findings-004.md` gained §4a, the argument for dtk and otk
+to read: no shared-specification change, because Part 12.2 already says the
+hook is handed the items and not through what; ztk's `on_close` was read at
+`pool.zig:127-130` and is still by pointer, an open divergence and not a
+recommendation. `-003` moved to `backup/`. `3tk-open-defects.md`'s `P6` row and
+section are marked ruled and built.
+
+**Verified**: `run-builds.sh` four builds green, 87 checks, 92 tests per build
+(the new `take()` test), 0 failures. `check-doc-loop.sh` 0 differing blocks,
+457 sentences, 456 found — the one miss is the pre-existing `inner.c3` module
+summary — 0 banned words. `run-sanitizers.sh` 3 passed, 0 failed, the hook path
+exercised on every run.
+
+**`grep -roiwE 'items?'` moved 365 to 366 in `ref`**, from `take()`'s two new
+sentences in the decisions entry; `3tk/src` is unchanged at 125, because
+`take()`'s doc block never says *item*.
+
+**3TK-50 is the only stage left** — the examples tree, independent of the fix.
+
 ## 2026-08-28 — the owner rules `P6`: the close hook takes the queue by value
 
 **A ruling, not a stage. No code changed.** Recorded here because the ruling and

design/secondary/lang/c3/3tk-log.md

@@ -1,6 +1,6 @@
 # 3tk — open defects
 
-**The working list for fixing the port.** Ten items — **seven fixed, one open,
+**The working list for fixing the port.** Ten items — **eight fixed, zero open,
 two closed**. One table, one section each: where it is, what is wrong, what the
 fix is, how to know it worked, and what state it is in.
 
@@ -33,7 +33,7 @@ are the good ones. Re-print before trusting them again — every fix moves them.
 
 | # | Where | What | Fix is | State |
 |---|---|---|---|---|
-| **P6** | `pool.c3:493`, `:563` | The pool can lose items and never know | **Ruled 2026-08-28: option 3.** `on_close` takes the queue **by value** — *I do not care what you did* — plus the wording | **ruled, built by 3TK-56.** [3tk-on-close-handoff-001.md](3tk-on-close-handoff-001.md) |
+| **P6** | `pool.c3:494`, `:564` | The pool can lose items and never know | **Ruled 2026-08-28: option 3.** `on_close` takes the queue **by value** — *I do not care what you did* — plus the wording | **ruled, built by 3TK-56.** [3tk-on-close-handoff-001.md](3tk-on-close-handoff-001.md) |
 | **Q5** | `mailbox.c3:124`, `pool.c3:253` | A release racing a call still in flight | Ruled 2026-08-28: stated, and checked | **fixed 2026-08-28, 3TK-53 + 3TK-54, closed by 3TK-55** |
 
 **Six of the ten are done, on 2026-08-27, in one stage.** Every mechanical and
@@ -47,8 +47,9 @@ and they have been overtaken** — the live ones are in
 pool, both 2026-08-28; 3TK-55 closed this row against the built code on the same
 day, after re-running all three scripts.
 
-**One remains open, and it is waiting on you.** `P6` needs one small ruling
-before code can be written. The ruling is stated in the section.
+**`P6` is the eighth, and it is done.** Ruled 2026-08-28, built 2026-08-30 by
+3TK-56 — [3tk-on-close-handoff-001.md](3tk-on-close-handoff-001.md) is its
+charter. Nothing remains open.
 
 **`P7` no longer needs one.** Reading `std::thread` showed the fault it warned
 about cannot occur. The branch is dead, it stays, and a plain `//` comment at
@@ -70,10 +71,11 @@ Only the open and deferred items have an order. The fixed ones are done.
 
 **Waiting on you.**
 
-- `P6` — one ruling, three answers, written out in its section. INTR 1 ranked it
-  High, and it is the last of the three it put there.
+- Nothing. `P6` was the last one, ruled 2026-08-28 and built 2026-08-30 by
+  3TK-56.
 
-**Nothing is waiting on a stage any more.** `Q5` was the last one, and it ran.
+**Nothing is waiting on a stage any more.** `Q5` and `P6` were the last two,
+and both ran.
 
 **The dependency that is not obvious, and it is now discharged.**
 
@@ -112,9 +114,9 @@ every number in an earlier version of this table, is now wrong.**
 **Three sites.**
 
 ```
-pool.c3:493     the close hook, called with what was left after a race
-pool.c3:563     the close hook, called with everything the pool held
-pool.c3:526-527 one item with an identity the pool does not recognize  -- RULED 2026-08-27
+pool.c3:494     the close hook, called with what was left after a race
+pool.c3:564     the close hook, called with everything the pool held
+pool.c3:527-528 one item with an identity the pool does not recognize  -- RULED 2026-08-27
 ```
 
 **Re-printed 2026-08-28**, after the lifetime fix moved every line below
@@ -205,19 +207,33 @@ Documentation, not code — so that the silence is not read as a promise.
 writing code. 3 is cheapest and defensible. **1 and 2 combine** — a check while
 developing, a number afterwards.
 
-**State: RULED 2026-08-28 — option 3, and it is stronger than option 3 as
-written here.** `on_close` takes the queue **by value**, so the pool physically
-cannot observe what the hook did and the type says so; the comment and the
-reference say it in words. **Options 1 and 2 are closed for good, not deferred.**
-The ruling, the five defaults that go with it and the work list are in
-[3tk-on-close-handoff-001.md](3tk-on-close-handoff-001.md), which is **3TK-56's
+**State: RULED 2026-08-28, BUILT 2026-08-30 by 3TK-56 — option 3, and it is
+stronger than option 3 as written here.** `on_close` takes the queue **by
+value**, so the pool physically cannot observe what the hook did and the type
+says so; the comment and the reference say it in words. **Options 1 and 2 are
+closed for good, not deferred.** The ruling, the five defaults that go with it
+and the work list are in
+[3tk-on-close-handoff-001.md](3tk-on-close-handoff-001.md), which was **3TK-56's
 only input besides the status file.** The measurement behind it: a by-value
 struct parameter is an lvalue on c3c 0.8.3, so the hook's ergonomics are
 unchanged, and the caller's copy is untouched, so nothing is left to count.
+`InnerQueue.take()` (`../3tk/src/queue.c3`) is the O(1) move both call sites
+use.
 
-**Superseded state: open, waiting on your decision. Nothing blocked it** — the
-dependency on `Q5` is discharged, and all three options are on the table. See
-[Order](#order).
+**Superseded state: open, waiting on your decision.** Closed by 3TK-56.
+
+**What 3TK-56 moved outside `3tk/src`:**
+
+- `ref/3tk-decisions-004.md` — the `P6` entry, under `pool.c3` and `queue.c3`.
+  `-003` went to `backup/`.
+- `ref/3tk-reference-004.md` — the `on_close` signature and its doc block, and
+  `InnerQueue.take()` added to the queue's operation list. Edited in place, not
+  versioned: this stage's charter names the file directly.
+- `3tk-port-findings-004.md` — new §4a, the argument for dtk and otk. `-003`
+  went to `backup/`.
+- `3tk/test/t_queue.c3` — one positive test, `take_empties_the_source`. No
+  negative is possible: a hook that keeps items dereferences nothing, so no
+  build can notice.
 
 
 ## Refuted — claims that did not survive the code
@@ -310,27 +326,27 @@ row.
 
 ## After a fix
 
-**All the numbers below were re-measured on 2026-08-28, by 3TK-55, after 3TK-53
-and 3TK-54.**
+**All the numbers below were re-measured on 2026-08-30, by 3TK-56, after `P6`
+was built.**
 The one missing descriptor sentence is the pre-existing `inner.c3` module
-summary and is not new. `ref` moved 360 to 365 on 2026-08-27 because the reference and the decisions
-each took on new sentences that say *item*, and it has not moved since. `3tk/src`
-went 124 to 125 with the lifetime fix's new prose.
+summary and is not new. `ref` moved 365 to 366 with `take()`'s two new
+sentences in the decisions entry. `3tk/src` is unchanged at 125 — `take()`'s
+doc block does not say *item*.
 
 **Every mechanical item touches `3tk/src`, so the doc loop is owed.**
 `ref/3tk-doc-loop-003.md` is the procedure; `check-doc-loop.sh` says whether it
 is still owed.
 
-**`ref/3tk-decisions-003.md` is the current one**, and `-002` is in `backup/`.
-`P6` will add an entry when it is ruled; `P7` closed without one.
+**`ref/3tk-decisions-004.md` is the current one**, and `-003` is in `backup/`.
+`P6`'s entry is in it, under `pool.c3` and `queue.c3`; `P7` closed without one.
 
-**The numbers to re-measure**, all true on 2026-08-28 after 3TK-54:
+**The numbers to re-measure**, all true on 2026-08-30 after 3TK-56:
 
 ```
-./run-builds.sh        # four builds green, 87 checks, 91 tests per build, 0 failures
-./check-doc-loop.sh    # 0 differing blocks, 452 sentences, 451 found, 1 missing, 0 banned
+./run-builds.sh        # four builds green, 87 checks, 92 tests per build, 0 failures
+./check-doc-loop.sh    # 0 differing blocks, 457 sentences, 456 found, 1 missing, 0 banned
 ./run-sanitizers.sh    # 3 passed, 0 failed — thread on two builds, address on one
-grep -roiwE 'items?' 3tk/src ref        # 125 and 365
+grep -roiwE 'items?' 3tk/src ref        # 125 and 366
 ```
 
 **`run-sanitizers.sh` joined the list here** because the lifetime fix is a