Skip to content

Commit 83f4892

Browse files
committed
docs(dsh): round 28c — e2e results: rc.5 cortexkit#1 41842386831 133/1 (C3: peak at the last step after a 6th read pair, U turns at 66-72 %; 待验证, check unchanged), rc.5 cortexkit#2 41845836776 135/0/0; alpha.3 a3 41842392546 132/1 and a3b 41842397847 133/1, both Y2 only (the first step's busy probe catches the NEW process's own fresh lease 4 ms after it took over the dead holder's; takeover/noop/nothing left behind correct; check not relaxed, for the user). I1/I2/R10 PASS on all four. Two consecutive green NOT met at adb5ed8. alpha.3 compat table filled; soak seg 2 K1-K9 PASS cumulative; evidence c6-28c-c3-y2.md (no keys)
1 parent 95333e5 commit 83f4892

3 files changed

Lines changed: 75 additions & 4 deletions

File tree

‎docs/DSH-MAINTENANCE.md‎

Lines changed: 17 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -64,7 +64,16 @@ report). Evidence: `docs/dsh/e2e/evidence-round28/` (run table in its README).
6464
| 41840694149 | install | alpha.3 | 21985b4ef | failure | DSH plugin manager rejects the bundle (peer pin 0.2.1-rc.5) → exemption lane `adb5ed80d` | | | |
6565
| 41841943616 | install | alpha.3 (+allow-version) | adb5ed80d | A0 WARN, A1–A4 PASS | — (A0 WARN = pin-only doctor FAIL + exemption needed, by design) | | | |
6666
| 41841941636 | push checks | rc.5 | adb5ed80d | success | dsh-plugin 624/0 | | | |
67-
| R28C_RUNS |
67+
| 41842386831 | e2e rc.5 #1 (28c) | rc.5 | adb5ed80d | 133 / 1 / 1 (E2) | C3: peak at the last step (6th read pair, U turns at 66–72 %), 待验证, check unchanged (`evidence-round28/c6-28c-c3-y2.md`) | PASS | PASS | PASS |
68+
| 41845836776 | e2e rc.5 #2 (28c) | rc.5 | 95333e55d (docs only over adb5ed80d) | 135 / 0 / 0 | — (C3 PASS again) | PASS | PASS | PASS |
69+
| 41842392546 | e2e alpha.3 slot a3 (28c) | alpha.3 + allow-version | adb5ed80d | 132 / 1 / 2 (E2, T3) | Y2: first-step busy probe sees the NEW process's own fresh lease (check timing), Magic behaviour correct | PASS | PASS | PASS |
70+
| 41842397847 | e2e alpha.3 slot a3b (28c) | alpha.3 + allow-version | adb5ed80d | 133 / 1 / 1 (E2) | Y2: same as a3 | PASS | PASS | PASS |
71+
72+
"Two consecutive green" for round 28c (no FAIL besides known won't-fix upstream /
73+
DSH-native items) is **not met** at `adb5ed80d`: rc.5 #1 41842386831 has C3 FAIL
74+
(待验证, not a known upstream item), rc.5 #2 41845836776 is 135 / 0 / 0. The
75+
round 28b e2e #0 41839150646 @ `f6ce05a22` was also 0 FAIL, but on older code and
76+
not consecutive with #2. No third run was dispatched this round (budget).
6877

6978
Round 28 commits (adapter / scenarios / scripts / CI only, no core change; red test first):
7079
- `92c8f5127` RED: folded tool record must be one past-tense prose summary (fold-format).
@@ -1973,7 +1982,13 @@ script), `MAGIC_CONTEXT_DEBUG_HISTORIAN` (1), `OPENCODE_SESSION` (random).
19731982
| plugin manager install / boot | rejects the bundle: peers pin 0.2.1-rc.5 exactly | install 41840694149 failure | by design (DSH); CI grants `dsh plugin allow-version … --accept-risk` off-pin (`adb5ed80d`); users on alpha.3 would need the same exemption until peers change |
19741983
| doctor | exact-version pin FAILs (only that line) | 41841943616 A0 WARN `doctorPin=1` | expected off-pin; tolerated in CI only |
19751984
| install → upgrade → disable → roll back | PASS | 41841943616 A1–A4 PASS | — |
1976-
| real-model e2e | A3_E2E_RESULT | A3_E2E_RUNS | A3_E2E_DISPO |
1985+
| real-model e2e (135 judged) | 132–133 PASS, **1 FAIL each (Y2)**, I1/I2/R10 PASS | 41842392546, 41842397847 vs rc.5 41839150646 / 41842386831 | Y2 已复现 2/2 on alpha.3, 0 on rc.5: the first step's busy probe catches the new process's own fresh lease 4 ms after it took over the dead one; takeover, noop passes, nothing left behind all correct. Check timing assumption; not relaxed — pending user. Every other check matches rc.5 (T3 PASS/WARN and E2 WARN vary on both) |
1986+
| other e2e checks | same as rc.5 | same runs | no new FAIL other than Y2 |
1987+
- **28c e2e findings (no code change; checks not relaxed):** Y2 FAIL on
1988+
alpha.3 2/2 (check timing: first-step busy probe sees the new process's own
1989+
fresh lease), C3 FAIL once on rc.5 (peak at the last step), both in
1990+
`evidence-round28/c6-28c-c3-y2.md`. Neither is an adapter fault on the
1991+
evidence; both are listed for the user.
19771992
- **Fixed this round (adapter):** open-lock crash residues (`bb86b923c`);
19781993
re-own with ≥ 2 stale owners (`8ea0321a8`); silent catch in context plane
19791994
(`952a3970a`); alpha.3 preset include (`bdeb8bfa0`); test flake

‎docs/dsh/e2e/evidence-round28/README.md‎

Lines changed: 6 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -26,8 +26,11 @@ credentials.
2626
| 41841161491 | dsh-ci checks (push) | rc.5 | 5b3ae0917 | success | | | | |
2727
| 41841941636 | dsh-ci checks (push) | rc.5 | adb5ed80d | success | | | | dsh-plugin **624**/0 |
2828
| 41841943616 | dsh-install | 0.2.2-alpha.3 + allow-version | adb5ed80d | A0 WARN, A1–A4 PASS | | | | A0 WARN = only doctor's version pin failed; exemption needed (by design) |
29-
R28C_RUNS
30-
| 41775904921 | dsh-soak (12 h, 4 × 180 min) | rc.5 | 901fb7874 (round 26 code) | SOAK_STATUS | | | | |
29+
| 41842386831 | dsh-ci e2e rc.5 #1 (28c) | rc.5 | adb5ed80d | 133 / 1 FAIL (C3) / 1 WARN (E2) | PASS | PASS | PASS | C3 peak at the last step, 待验证 (`c6-28c-c3-y2.md`) |
30+
| 41845836776 | dsh-ci e2e rc.5 #2 (28c) | rc.5 | 95333e55d | 135 / 0 / 0 | PASS | PASS | PASS | — (C3 PASS again) |
31+
| 41842392546 | dsh-ci e2e (slot a3) | 0.2.2-alpha.3 + allow-version | adb5ed80d | 132 / 1 FAIL (Y2) / 2 WARN (E2, T3) | PASS | PASS | PASS | Y2 check timing (`c6-28c-c3-y2.md`) |
32+
| 41842397847 | dsh-ci e2e (slot a3b) | 0.2.2-alpha.3 + allow-version | adb5ed80d | 133 / 1 FAIL (Y2) / 1 WARN (E2) | PASS | PASS | PASS | Y2 same |
33+
| 41775904921 | dsh-soak (12 h, 4 × 180 min) | rc.5 | 901fb7874 (round 26 code) | seg 1+2 success, K1–K9 PASS cumulative (6681/6681 mock, real 12/12); seg 3 running, seg 4 queued at 22:25 CST | | | | |
3134

3235
## Files
3336

@@ -36,4 +39,5 @@ R28C_RUNS
3639
- `c3-reown-multi-owner-*` — re-own with ≥ 2 stale owners red/green.
3740
- `c4-silent-catch-*` — context plane silent catch red/green.
3841
- `c5-r2-r4-r7-model-selfdrop.md` — R2/R4/R7 analysis (model ctx_reduce against the prompt).
42+
- `c6-28c-c3-y2.md` — 28c: C3 on rc.5 #1 and Y2 on alpha.3, offline re-check.
3943
- `b1-preset-include-commit-*` — alpha.3 preset include commit() red/green + repro.
Lines changed: 52 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,52 @@
1+
# Round 28c — C3 (rc.5 #1) and Y2 (alpha.3 ×2)
2+
3+
No keys here (redacted CI artifacts only). Offline re-check: `ci-check.py check` on
4+
the downloaded `dsh-e2e-<run>` artifact, with the 300-char detail cut disabled.
5+
6+
## Y2 on DSH 0.2.2-alpha.3 — 已复现 2/2 (41842392546, 41842397847); 0 on rc.5
7+
8+
Y2 asserts that in y-2 (new process after kill -9 during a historian hold) the
9+
**first step's** busy probe reads false. On alpha.3 the first step's status dump
10+
(41842392546) is:
11+
12+
killedPid 4789 (y-1) -> y-2 first step: historianBusy=true,
13+
leases=[{holder dsh-historian:…:4847:a8cca8e6, owner_pid 4847, expires_at +300 s}]
14+
lease acquired at …153217, step status at …153221 (4 ms later)
15+
16+
pid 4847 is the NEW y-2 process, not the dead holder: the dead lease was taken
17+
over immediately (no 5-min TTL wait), both y-2 historian passes ran `noop`, and
18+
after the turn there is no live lease and `compartment_in_progress = 0`. Every
19+
other Y2 condition holds; `still_dead` is empty. On rc.5 (41839150646) the same
20+
pre-step trigger fires, but by the step's status dump the noop pass has already
21+
released the lease (`leases: []`, busy false). The y-2 log order is identical on
22+
both versions (trigger → native check → step_start).
23+
24+
Classification: **timing assumption of the Y2 check** (the probe samples while
25+
the new process's own pass still holds its fresh lease); Magic's behaviour
26+
(dead-holder takeover, no wait, nothing left behind) is correct on alpha.3.
27+
Why alpha.3 starts the step sooner is **static inference** (faster pre-step on
28+
alpha.3), not reproduced locally. Not changed: the check is not relaxed (user
29+
rule); whether Y2 should accept a busy probe whose only lease belongs to the
30+
live new process is listed for the user.
31+
32+
## C3 on rc.5 #1 (41842386831) — first C3 FAIL in the 15 runs re-checked from round 27 on
33+
34+
C3: the lowest prompt-token value **after the global peak** must be below it.
35+
36+
cmp-02 [10313, 15093, 19008, 19084] cmp-03 [19301, 24067, 19235, 19385]
37+
cmp-04 [19578, 23206, 23364, 23729] cmp-05 [23923, 22564]
38+
cmp-06 [21115, 27264, 27777] cmp-q [26262]
39+
use-search [26271, 26504] use-memory [27460, 27536] use-expand [28351, 28638]
40+
peak 28638 (use-expand, the LAST step) -> nothing after it -> FAIL
41+
42+
Context did drop (cmp-03 24067 → 19235, cmp-05 → 22564, cmp-06 21115), the
43+
historian fired (C1), wrote 7 contiguous compartments (C2/C6), the codeword was
44+
answered (C4/C8). Compartments first appeared after cmp-05 (the e2e
45+
corrupt-once injection + repair retry landed there; on 41839150646 after cmp-04),
46+
so the scenario used its 6th read pair; the two newest full-file reads stayed in
47+
the protected tail and the U turns ran at 66–72 %, each adding a little, so the
48+
peak is the final step. Classification: **待验证** — most likely scenario
49+
variance plus a check that assumes the peak precedes compression; no adapter
50+
fault found in the logs (I2 PASS, no re-reads; the model once misread a
51+
one-call fold note as "both files ran" and then read the second file, which had
52+
not been read yet — correct new work). Check unchanged.

0 commit comments

Comments
 (0)