rdmsm4x - build - RTTy 524 Standard trust chain re-pinned 157-188 - claude - TASK-20260925-15 - RTTy - 2026-09-25 1933 EDT

HANDBACK — TASK-20260925-15 RTTy 524: register 8 Standard sources + re-pin the Standard trust chain

[2026-09-25 19:31:54 EDT · rdmsm4x] · claude@rdmsm4x/97f5c581 (headless, dispatched by grok@rdmsm4x/fleet-wd). Work ran from 19:28:25 to 19:32 EDT.

Status: DONE on grok/524-standard-trust. Not merged to main or grok/524-integrate. No version bump, no qa_release.sh or build_and_run.sh, no App Store Connect, no notarize. The 522 freeze (TASK-20260925-01) still holds.

Where it is

Worktree: /Users/richh/dev/_scratch/rtty-524-standard-trust-20260925. Branch grok/524-standard-trust, forked from grok/524-integrate a5c3398.
One commit: 190a271eeab970e4d9afba8073d140e2e7f9d566, pushed to backup without force. git ls-remote backup returns the same SHA. Nothing was pushed to origin or fleet.
The commit touches the same 9 files as the pattern commit 9d6785d (+141 / −22), staged by explicit path. The worktree is clean afterwards.

What was done (the procedure in `docs/release/RELEASE-PROCEDURE-RTTy.md` §1, steps 1–8)

Step	Result
1–2 `project.yml` + `StandardSources.txt`	3 RTTy files were added to both files at the same relative position: `Stores/ObservationStore+FriendlyVerdict.swift` and `Stores/ObservationStore+UsageProfile.swift` after `ObservationStore.swift`, and `Views/UsageProfileSettingsSection.swift` after `SettingsView.swift`. StandardSources.txt went from 75 to 78 lines, and the project.yml app-target sequence diffs identical to it. The 5 LatencyCore files were added to the project.yml LatencyCore target in alphabetical position (`Profiles/CustomServiceProfileManager`, `Profiles/DiscoveredProbeKey`, `Profiles/UsageProfile`, `Verdict/ApplicationCheckSummary`, `Verdict/NetworkVerdict`). LatencyCore is taken whole, so it has no StandardSources.txt entry.
3 xcodegen 2.46.0	rc 0. The pbxproj gained 41 lines (4 references per file). No other file was touched: no Info.plist or scheme changes.
4 pins	`EXPECTED_COMPILER_INPUT_COUNT` 149 → 157 and `EXPECTED_BUILD_INPUT_COUNT` 180 → 188, in both `standard_project_graph.py` and `standard_archive_validator.py`, each with a ledger entry that gives the source commit of every file (1481739, 390ad31, 656b44a, e8c89d1, 8ecd722). `standard_candidate_closure_test.py` pins changed to `(79, 69, 1, 8)` / 157.
5 source policy	`standard_source_policy.py --path <f>` returned rc 0 for all 8 files. As a negative control, a scratch file containing `Process()` returned rc 1 ("Foundation Process").
6 render-graph	Rendered into scratch and diffed. The only changes are the 8 files, the counts, and the hashes of project.yml, the pbxproj, StandardSources.txt and the template. The Release.xcconfig, scheme and workspace hashes are unchanged, so there is no version drift. The template pbxproj and graph were copied in.
7 verifier bundle	`standard_verifier_bundle.py render` output was 4330 bytes and 26 entries. A second render and a third render were byte-identical: a fixed point. Bundle sha256 is `9016178d3f7e404eaf264d3bce6ad2b587fdabfd246deee54410d5d8d312bb6d`.

Tests — rc and counts, not "green"

Test	Before (a5c3398)	After (190a271)
`python3 Tests/ScriptTests/standard_source_reference_closure_test.py`	rc 1 (5 LatencyCore files missing)	rc 0: 78 allowlisted, 13 Direct-only, 69 LatencyCore registered
`python3 Tests/ScriptTests/standard_pbx_source_closure_test.py`	rc 1	rc 0: every Standard PBX Sources phase matches its projection
`python3 Tests/ScriptTests/standard_candidate_closure_test.py` (materializes HEAD)	rc 1 (got 76/69/1/8)	rc 0: 157 compiler inputs
`standard_frontend_input_closure_test.py`, `standard_project_test.sh`, `standard_archive_manifest_test.py`, `standard_launch_capability_policy_test.py`, `standard_source_policy_adversarial_test.py` (67 caught / 9 controls), `standard_source_policy_alias_test.py`, `standard_snapshot_lifecycle_test.py`, `standard_app_intents_metadata_test.py`	—	rc 0 (8 of 8)
`git diff --check`	—	clean

Three tests fail both before and after this change. None was caused by it:

standard_trust_entry_policy_test.py and archive_standard_policy_test.sh both report "installed stable Python interpreter bytes differ from the trusted graph", with rc 1 on an export of the base a5c3398 too. The graph pins the /usr/bin/python3 real path from the Xcode 26.6 toolchain, and this host currently runs Xcode 27. This is environmental, and the release runbook's 26.6 swap (§0) is what satisfies it. This commit did not change the graph's toolchain block.
mobile_pbx_source_closure_test.py fails because Apps/RTTyMobile/App/RTTyCloudKitRoundTripProbe.swift is not in the reviewed Mobile entry points. That file was added by 2cd3c60, which is already on main. This commit's diff against a5c3398 under Apps/RTTyMobile is empty. It belongs to whoever owns the iOS debug probe, not to this ticket.

Not done (out of scope or release-time work)

Launch-capability policy re-render (runbook §2). It needs a 26.6 archive of the candidate. The 8 new files add SwiftUI surfaces (the Settings section and the verdict), so expect new symbols. Grep the delta for Process|NSTask|posix_spawn|popen. This is a release-time step, done with the bump after the freeze lifts.
No Standard xcodebuild compile was run: no gate and no archive, per the hard rules. The closure tests prove registration and closure. They do not prove the Store target compiles under Xcode 26.6.
Merging grok/524-standard-trust into grok/524-integrate / main is the lead's job. It is a single commit on top of a5c3398, so it fast-forwards onto integrate.

Scratch left behind (tmp, not repo)