Repository navigation
Conversation
(cherry picked from commit eec419948e1ea132bd233741855d9037e9c8d1d4)
(cherry picked from commit 9ffe2f4)
Development Debug and Release both require iOS 15.1 and use the legacy armv7 requirement. Match the existing production arm64 capability and Apple guidance for 64-bit apps. Add a source-bound regression for the active configurations. This is a source-only proposal; simulator launch causality is not established.
The actual iOS attempt passed driver readiness and verified the original installed app but received a generic FrontBoard launch denial before any first screen. The existing app-process-only log query cannot see an OS denial before app startup. Read system launch-service logs on the same owned simulator and original startup window, projecting only fixed services, allowlisted domain/code pairs and finite markers. Keep raw output in a closed private temporary file; observer failure cannot replace the startup failure or grant acceptance. No rerun is selected by this commit.
|
One combined native validation is selected at exact head 5bb116a, tree538974beb448a6d64058fdda384e8c225fc0e42e: https://github.com/KeeperCommunity/bitcoin-keeper/actions/runs/40161613081 (attempt1). iOS job121074185201 and Android job121074185388 are actually running on independent hosted machines. The source remains fixed while they run. Root and independent exact-source integration reviews are CLEAR. All previously reviewed app/frozen-dependency/driver changes are preserved; only diagnostic platform scheduling is parallel. The initial targeted Node integration failure identified one stale serial-scheduling assertion; the corrected final64 controls pass with no skips. The449 Python checks at parent973c are byte-bound to unchanged Python/application source at this head. Three actionlint checks and nine dependency/release-input fences pass. This selection tests actual native library initialization and app startup/first-wallet behavior; source/JVM tests do not provide that acceptance. The scoped backup software proof is also running against the same app source through the existing authorized private route. Previous failures on other heads remain preserved and no pass transfers between sources. No additional identical-source retry is authorized. No R4/main R5 source promotion, protected installation, production signing, store upload, distribution or deployment occurred. |
Keeper's cloud validation is blocked by unresolved startup/first-wallet failures. Android billing has a source-confirmed missing native package registration: the pinned CLI does not recognize IAP's BaseReactPackage subclass, so the already-linked module's initializer is omitted. This candidate registers it only when absent and owns optional billing init/cleanup promise rejections while retaining adapter error reporting.
The iOS validation host now matches the pinned Maestro vendor's explicit macos-15-intel runner. Earlier unsuffixed macos-15 runs actually selected ARM64 hosts; the most recent run40158447604 failed in driver setup before Keeper built. That remains a recorded failure and does not establish the cause of earlier app launch denial. The candidate retains the reviewed development CPU capability correction and captures existing bounded driver failure facts before owned simulator cleanup.
Exact head: 5bb116a; tree 538974beb448a6d64058fdda384e8c225fc0e42e. This is a NEW isolated candidate combining reviewed iOS ed71/8657/2d527 and Android1f01 descendants. Original R4, main R5 integration #7068, and diagnostic #7158/#7162 sources remain unchanged. No private R5.2 or private security implementation is included.
Only the opt-in diagnostic workflow's platform dependency was removed so separate hosted iOS and Android machines can execute concurrently on this same source. Both platform results remain required; draft/foreign-source/live-head/opt-in/modifier/R4-exclusion guards, read-only permissions and noncancellation remain. Canonical and release workflow sequencing/authority are unchanged.
Validation: exact compiled MainApplication/IAP-package regression and independent source reviews establish the registration defect and guard behavior. The combined source passes the existing 449 Python controls (parent973c, all Python/application bytes unchanged at final head), 64 focused actual-source Node controls with zero skips, three actionlint checks, frozen dependency/release-input fences and source/whitespace checks. The first Node integration run exposed one stale serial-scheduling assertion; its failure is preserved and the final corrected policy expectation passes. Shellcheck/pyflakes are unavailable. Hosted native/first-wallet acceptance is not established by those checks.
Previous actual outcomes remain source-qualified: Android0facc/run40111205437 built then failed Create-passcode before Skip; Android9ffe/run40105347015 passed startup/persistence/workers then failed first product StartNew; iOSb4/run40110349118 built and reached driver readiness then OS denied launch; iOSed71/run40158447604 failed in driver setup before build. No original failure or cross-head pass is reassigned.
Codex local handoff
State: NEEDS_LOCAL
PR: this isolated R5 cloud completion candidate
Base: sprint
Head branch: codex/r5-cloud-app-completion-20261025
Head commit: 5bb116a
Goal and acceptance criteria: genuine iOS and Android startup, native persistence/worker probes and full first-wallet/product scenarios on the same exact source.
Changes made: reviewed billing promise ownership, guarded Android IAP package registration, explicit iOS Intel host, development capability correction, existing failure capture before cleanup, and parallel diagnostic scheduling.
Cloud checks: source/regression checks and attributed independent reviews PASS as scoped above. Final current-source native execution NOT RUN at PR creation.
Required local checks: IOS, ANDROID, HARDWARE
Local checks requested: hosted iOS and Android native execution on isolated machines; separately scoped physical phone/signer checks remain distinct. No personal Mac or release signing/store credentials are needed for the diagnostic app jobs.
Risks / known gaps: actual Android JNI registration, first-screen/first-wallet outcomes and iOS host correction effectiveness remain unverified; backend proof/canonical installation/5.2/security/physical/release operations remain separately scoped. No merge, production signing, upload, distribution or deployment.