diff --git a/CONTEXT.md b/CONTEXT.md index 882185c6..41620b5b 100644 --- a/CONTEXT.md +++ b/CONTEXT.md @@ -1,5 +1,7 @@ # CONTEXT.md — Project Memory +> **2026-11-03 (day, attended) — v0.350.0 RELEASED, the global floor since 08:18 CET (MinAgent 0.158.1).** Image `sha256:9ee53908…` from `5c336d8`. Traefik v3.7.17 (GHSA-xv5x), R-1347, R-1348, R-1339, R-1351, R-1323. Golden 0.350.0 vouched 10:18:10 CET. +> > **2026-10-29 (evening, attended) — v0.349.0 RELEASED, the global floor since 19:24 CET (MinAgent 0.158.1).** Image > `sha256:c5341883…` from `94049a7`. D65 = A: `internal/busymark` keeps `/ctl/busy` fresh while the controller's > own work runs (the agent refuses „Reboot guest" meanwhile); the dashboard shows „A Felhom ügyfélszolgálata HH:MM-kor diff --git a/REPORT.md b/REPORT.md index dad626b0..b5b2e589 100644 --- a/REPORT.md +++ b/REPORT.md @@ -1,16 +1,19 @@ -# REPORT — 2026-11-02 night: five small fixes and the Traefik pin, commit only (no release) +# REPORT — v0.350.0 released as the global floor: Traefik v3.7.17 first (2026-11-03, attended) -**Plain words.** Nothing was released. The next controller release carries: Traefik v3.7.17, which closes a real hole on every box; a restored app is watched again; a failed put-back at start-up tells the operator; three false or noisy start-up lines go. +**Plain words first.** Every box runs 0.350.0 since 08:18 CET. It closes a real hole: a stranger with no password could fill a box's web front door until it stopped. It also carries five small fixes. -| Commit | CI (job id) | What | +| Commit | CI | What | |---|---|---| | `c4b6844` | 2551 green | Traefik v3.7.17 (GHSA-xv5x-242w-7hf7); R-1347, R-1348, R-1339, R-1351, R-1323 | -| `3764445` | 2560 green | R-1348: the mail names a restore, not an off-site run (seen on the live proof) | +| `3764445` | 2560 green | R-1348: the mail names a restore | +| `5c336d8` | 2568 green | v0.350.0 heading — the release build's commit. MinAgent 0.158.1 | -**Red-proofs.** Each `TestR13xx_*` was seen failing before its code existed, and each rule again by a mutant (10 of 10 red: `felhom.eu documentation/audits/security-2026-11-02/part3/`). +- rc.1 (`13e34d4`, `sha256:cbf4d42f…`) on 9202: walk green, sign-in works. +- 0.350.0: `sha256:9ee53908e6d95f0340c41bf4eaba954280084a8e6927de59779440e3551df261`. +- Per-box floors 08:04:24 CET: all three moved within 45 s, walk green, 26 public names answer, `https_only_check.py` green, sign-in works on Tester 1 and demo-hp (not driven on the lived-in N100). +- One restore on Tester 1 on the release build: clean. +- Global floor 08:18:16 CET, its own save; the 30-minute watch stayed healthy, 0 ERROR or held lines at the hub. -**Live, scratch 9202, a local test image of `c4b6844`** (`felhom.eu audits/security-2026-11-02/part3/README.md`): R-1347 and R-1348 proven through staged real states; R-1339 nine false lines → none; Traefik v3.7.17 ran through three boots. **R-1351: no Conflict on six boots of either image, so the live boots do not tell them apart; the proof is the test.** R-1323: not seen live. +**The load test on 9202:** v3.7.16 18.7 → 79.0 MiB after 920 NTLM connections, not released; v3.7.17 flat near 25 MiB for 25,116. -**R-1339, the row's own fix was wrong.** „A sub-folder of a mounted drive is not skipped" would write a `userdata/` inside the app's folder and recreate the file manager. The behaviour stays; only the false line goes. - -**Next:** on the operator's word, the release: controller v0.350.0, MinAgent unchanged (0.158.1). +**Test proof only:** R-1351 and R-1323 (their conditions did not fire). R-1347 and R-1348 were proven on the test image of the same code, not on the release build. Evidence: felhom.eu `documentation/audits/release-2026-11-03/part1/`.