docs(wiki): 2026-11-15 evening status refresh - held-window release carrier deterministic (c772), slot-34/SA[34] settled (0x265), call order is the blocker; eng64 chains the release behind the pack carrier; STATE-CURRENT snapshot

1 files changed, 52 insertions, 0 deletions

diff --git a/STATE-CURRENT.md b/STATE-CURRENT.md
index 3cf5c30..0414c56 100644
--- a/STATE-CURRENT.md
+++ b/STATE-CURRENT.md
@@ -1,5 +1,57 @@
 # Open-IOV Current State
 
+> **Snapshot: 2026-11-15 ~16:35 (SESSION 157 continued; HEAD `6aeb20c7` ==
+> `vgpu/master`; the working tree carries only this docs refresh; supersedes
+> the 2026-11-15 ~13:55 entry below).**
+> The graphics front moved from "the DS stage image is live" to "the release
+> path is deterministic; the blocker is the draw call's **order**". The
+> **held-call release path is now host-driven**: a saved-window release
+> carrier snapshots the held window at `MME-CALL-HOLD` and re-submits it with
+> the calls released once the pack is resident — proved end to end on c772
+> (`MME-CARRY-SAVE win=240 held=68` -> `MME-CARRY-FIRE mode=release` ->
+> synthetic `GR LAUNCH push=0x6d6d6500` -> `MME-PACK-REARM` gen=5 ->
+> `DS-FAST 0x84000420` + `DS-IMAGE st0=st1=020a0481` ->
+> `MME-CARRY-COMPLETE`). The slot-34 `DRAW_IAB` question is settled: the call
+> **is** in the saved 728-B window and its host entry is correct
+> (`MME-CALL-TARGET slot=34 method=0x3910 client=0x265 golden=0x265
+> used=0x265`; `MME-CALL-CENSUS ... calls=11 slot34=1`), and the SA[34] golden
+> fill lands `entry=0x265` (75-word IR body) without overwriting a live client
+> entry. The **remaining blocker is call order**, not window choice or SA
+> residency: c780's release replay ran `mac=0x33/0x39/0x0` because the 728-B
+> state window places the slot-34 CALL at `off=0x340`, after the `0x33/0x39`
+> setup; the last `0x39` latches DS `0x84000420` and the MFE never reaches
+> CALL(34). c716's 324-B DRAW window has CALL(34) first (off `0x1c0`) and ran
+> `mac=0x22`. A release-window reorder (`a03a68ea`: `[head binds][slot-34
+> run][rest]`, receipts `MME-DRAW-PLAN`/`MME-DRAW-MOVE`) is landed but **not
+> yet exercised** — the c781/c782 boots never completed the pack. Pack
+> completion itself is still a per-boot lottery with three observed classes: a
+> poisoned pack-part (c777, `badf1201` / `CARRY-CANCEL ctx-wipe`), no pump pass
+> after `MME-CARRY-ARM` (c779), and carry-armed-never-fired (c781/c782; one
+> `CANCEL chan-ctx-reset`). eng64 (`6aeb20c7`, landed) chains the release
+> behind a successful pack carrier and names a stalled carrier's gate
+> (`MME-CARRY-WAIT reason=...`), addressing the c779 no-pump-pass class;
+> validation pending. The DS stage image and the authored VTG SPH remain
+> live and stable (`st0=st1=020a0481 code=ff007321`; DS `0x84000420` = TI +
+> SPH5/10/18), and the DS `0x84000420` bits are the raster gate after the
+> draw-issue call; stores are still 0 (`NOWRITE-CONFIRMED ... wrote=0`). The
+> latest batch (c783-c785) was shallow — X green, `gl_frames` 3/3 black,
+> GZDoom rc=255, no chain — i.e. the X/GL start lottery remains the run-rate
+> limiter.
+>
+> CUDA front unchanged (backgrounded while graphics holds the device; its
+> 10-30 status stands): `41343c63` green (`victory_mail` `VADD bad=0` +
+> `STATUS OK`; `vecadd_loop` 100 launches PASS); the fresh-boot battery's red
+> class is D2H-stage with release-discipline fixes staged; the SESSION 143
+> Vulkan fence regression is not re-measured.
+>
+> Live at check (read-only): host uptime 6 h 32 m (fresh boot 2026-11-15
+> ~10:03); `vgsp` loaded (refcount 0) + mdev
+> `6320cb2a-5e86-4e37-a21b-3defbff5bbec` + QEMU cycle c786 (PID 76485, 12 GiB
+> file-backed guest, kvmfr ivshmem, since 16:34); `nvidia-smi` reports no GPU
+> (the claim is held); pstore empty; no Xid/Oops; no reboot during the check.
+> The open-iov.org wiki was refreshed alongside this snapshot (11 pages, revs
+> 24177-24187; 68 pages audited, 0 problems).
+>
 > **Snapshot: 2026-11-15 ~13:55 (SESSION 157 continued; HEAD `692a7eb0` ==
 > `vgpu/master`; the working tree carries only the campaign parameter file
 > `tools/guest-diag/params-r1624.txt`; supersedes the 2026-11-14 ~10:55 entry