Skip to content

docs: record release 0.1.0-6e7e065-94d85365 as DG-1's qualified release - #34

Merged
novelKR merged 2 commits into
mainfrom
claude/project-thread-af6ntg
Oct 13, 2026
Merged

novelKR merged 2 commits into
mainfrom
claude/project-thread-af6ntg

Conversation

@novelKR

@novelKR novelKR commented Oct 13, 2026 •

Copy link
Copy Markdown
Owner

Records the SLO qualification of release 0.1.0-6e7e065-94d85365, measured on the qualification host on 2026-10-13 and promoted, as DG-1's qualified release. This is step 5 of CodeSpace's CSRG-U6 plan: U6 qualifies CS-RG against a DG-1-qualified release under the same policy.

Why

The ledger names 0.1.0-5daee5d-b3fa569e, which protocol-2 qualified on 2026-09-26. A qualification holds only for its release, host and policy, and the release and the policy have both changed since. On 2026-10-12 the qualification host was upgraded to 0.1.0-6e7e065-94d85365, built from source 6e7e065 (#27's merge), the client pin CodeSpace links (novelKR/CodeSpace#89). CS-RG then added CodeSpace's codespace consumer to the host's policy: a control service with one instance and a reservation of 150 mCPU, 512 MiB and 40 tasks. The owner had the installed release measured again before U6's qualification runs, and ran the reviewed command in a terminal on the host on 2026-10-13.

Measurement

scripts/measure.py macos ran from a clean worktree at ea98d7f, whose measure.py and fixture are protocol-2's. A rehearsal (04:18–04:23Z) was inconclusive by design, with no other reason. The full protocol then ran from 04:23:42Z to 08:31:29Z.

Item Result
Verdict qualified, exit 0: cold and warm, three repetitions each, all passed, with no reason in any interval
Input to next paint p99 at most 56 ms (target 100 ms); no input lost
Frames largest gap 18.8 ms; no stall
Status p99 at most 46.2 ms (target 500 ms)
Termination acknowledgement p99 at most 167.9 ms in one idle interval and at most 77.1 ms in the others (target 1,000 ms); every termination effective
Load 1,370 runs, all succeeded; no connection loss, uncertain execution or duplicate launch
Service pid 23023 before and after every repetition; all 2,939 host pressure samples normal
Run header source 6e7e065, clean; MANIFEST.json SHA-256 4535b52c…d0df; host.toml SHA-256 aa40c637…f3a9, configuration fingerprint 8cbb019b…c6f; harness ea98d7f, not dirty; summary.json SHA-256 7a4dfa12…2e9c
Promotion promoted: true after verifying that the service runs the release; record qualifications/0.1.0-6e7e065-94d85365.json, SHA-256 fd4fed31…17a4
Evidence sealed on the host: csrg-u6-dg1-measure-2026-10-13 (1,469 files, MANIFEST.json SHA-256 b4c07233…380c) and csrg-u6-dg1-rehearsal-2026-10-13 (86 files, 9ce9cb1a…26c1); leak scan empty

Changes

File Change
milestones.json DG-1's qualified_release is 0.1.0-6e7e065-94d85365; its qualification stays macos-qualified
README.md, docs/milestones.md DG-1's qualified release; the milestone row also dates both measurements
docs/planning/README.md, docs/ko/planning/README.md The status sentence and DG-1's row name the new release
docs/planning/verification.md, docs/ko/planning/verification.md V-DG1-SLO names the new release and its policy, and keeps protocol-2's
docs/planning/milestones/DG-1.md, docs/ko/planning/milestones/DG-1.md The header names the new release; the re-measurement's result follows protocol-2's
docs/planning/consumer-readiness.md, docs/ko/planning/consumer-readiness.md R2 and RS are permitted with the new release
docs/planning/milestones/CS-RG.md, docs/ko/planning/milestones/CS-RG.md The dated current-state note, now 2026-10-13, records the re-measurement and that U6 qualifies against that release

Korean counterparts reviewed; planning-readme, planning-verification, planning-milestones-dg-1, planning-consumer-readiness and planning-milestones-cs-rg hashes recorded with scripts/check_docs.py record.

What does not change

  • Protocol-2's result for 0.1.0-5daee5d-b3fa569e stays recorded, and no historical text is rewritten: the CS-RG plan's baseline entry and CSRG-C01's candidate, the CodeSpace integration specification and the handoff documents keep the release they named.
  • CS-RG is not marked implemented or qualified, and U6 stays active. Linux stays not-run.
  • No product code, contract, pin, dependency, CI policy or test changes.

Verification

At 4a1d6b3, based on main at ea98d7f:

  • python3 scripts/ci_plan.py --base origin/main --head HEAD: profile affected, 14 normative paths, stages whitespace, source-contract and documentation, no Rust.
  • python3 scripts/validate.py --stages whitespace,source-contract,documentation --diff-base origin/main: all three passed.
  • python3 scripts/check_docs.py: passed, 18 reviewed pairs, 48 work units, 25 logical PR groups.
  • python3 -m unittest discover -s scripts -p 'test_ci_*.py': 95 tests passed.

Rollback

Revert the merge (git revert -m 1 <merge commit>). The ledger then names 0.1.0-5daee5d-b3fa569e again, which the host no longer runs. Revoking the qualification itself is a separate step on the host: remove the record qualifications/0.1.0-6e7e065-94d85365.json, which nothing reads to admit work.

On 2026-10-13 the owner ran scripts/measure.py macos from a clean
worktree at ea98d7f against the installed release
0.1.0-6e7e065-94d85365, built from 6e7e065, under the policy that adds
CodeSpace's codespace consumer (CSRG-U6). The rehearsal was
inconclusive only for its short durations; the full protocol passed all
three repetitions of both combinations, and promote recorded the
release after verifying that the service runs it.

Point the ledger's qualified release, the README, the milestone table,
the planning index, the verification row, DG-1's header and the
consumer-readiness gate at the new release, and add the re-measurement
to CS-RG's dated current-state note, English and Korean. Protocol-2's
result for 0.1.0-5daee5d-b3fa569e stays recorded. Korean counterparts
reviewed; planning-readme, planning-verification,
planning-milestones-dg-1, planning-consumer-readiness and
planning-milestones-cs-rg hashes recorded.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The re-measurement ran DG-1's SLO protocol with harness ea98d7f, whose
measure.py and fixture are protocol-2's, against release
0.1.0-6e7e065-94d85365 under the policy with CodeSpace's codespace
consumer, from 04:23:42Z to 08:31:29Z. Both combinations passed all
three repetitions: input to next paint p99 at most 56 ms, status p99 at
most 46.2 ms, termination acknowledgement p99 at most 167.9 ms in one
idle interval and at most 77.1 ms elsewhere, no frame stall, and 1,370
load runs without connection loss, uncertain execution or duplicate
launch. Record it beside protocol-2's result, English and Korean;
planning-milestones-dg-1 hashes recorded.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants