local-first ios domain inspection toolkit. dns, tls, rdap, audit. http://git.krz.sh/krz/domain-dig.git/atom?h=v5.0.1 2026-07-25T03:52:53+00:00 2026-07-25T03:52:53+00:00 Christian Cleberg hello@cleberg.net 2026-07-25T00:27:24+00:00 urn:sha1:a7d154e1db9fd9904b2255db889144ad15aac2f2 Stands up the DomainDigTests unit-test target the v5.0.0 roadmap flags as the mandatory first move before decomposing the god-files and locking external contracts. The project had no XCTest unit target — only the DomainDigUITests accessibility suite. - New DomainDigTests target (unit_test_bundle), hosted by the app so @testable import DomainDig links. Mirrors the UITests build settings (SWIFT_VERSION 5.0 + MainActor default isolation) to avoid the XCTest override-isolation issue recorded for the Swift 6 targets. Wired into the shared DomainDig scheme's Test action, so it runs in CI and the pre-push audit hook automatically (both invoke the whole scheme). - SnapshotFixture builds the deep LookupSnapshot/DomainReport models through their real initializer and the DomainReportBuilder, exposing only the fields the tests vary. - 36 characterization tests across the four deterministic units the roadmap names: - DiffService: change classification, case/whitespace normalization, DNS record reorder-vs-change, summary phrasing, resolver context note, certificate-warning thresholds. - DomainReportBuilder: snapshot -> report field mapping, primary-IP and DNSSEC derivation, TLS status, partial-snapshot/validation passthrough. - DomainReportExporter: format dispatch, JSON round-trip (the machine contract), CSV/markdown/text/PDF structural invariants, timeline. - DomainDataPortabilityService: CSV round-trip and the merge/dedup semantics (case-insensitive collapse, OR-merged pin state, min-created/max-updated, recency sort) via an ephemeral UserDefaults suite. - build.yml comment updated to reflect the second test target. 2026-07-20T22:55:24+00:00 Christian Cleberg hello@cleberg.net 2026-07-20T22:21:03+00:00 urn:sha1:38723030eb0db58a9d62a45c7725861901309a05 Phase 0 of the accessibility pass: a regression guard that must exist before any of the remedial phases, so their acceptance criteria are enforced rather than asserted once by hand. - Fix the project-level IPHONEOS_DEPLOYMENT_TARGET, which was 26.2 while all three targets are 17.6. It was shadowed everywhere today, but any target added later would silently inherit it and drop iOS 17.6 support with no error. - Add a DomainDigUITests target running performAccessibilityAudit on the six primary screens, plus a sweep of every root screen at AccessibilityXXXL. Uses the existing DOMAIN_DIG_FORCE_PRO_PLUS debug argument so Pro-gated screens are reachable. - Findings are reported, not failed. The audit surfaces violations that exist today, so gating on them would block unrelated PRs until the whole pass lands. Enforcement is a committed constant, AccessibilityAuditHarness.enforcedAuditTypes, widened per audit type as each phase clears a category. - CI now runs xcodebuild test across two simulators. Audit coverage is not nested between OS versions: on Tracked Domains, iOS 18.6 reported 2 findings and iOS 27.0 reported 6 (including contrast and element-detection issues 18.6 never raised), while at accessibility text sizes the Dashboard produced a hit-region finding on 18.6 that 27.0 did not. - Simulator selection is now dynamic and floor-aware. The previous selector took the first iPhone from any runtime, which can resolve to a simulator below the deployment target where the app cannot install. Baseline on iOS 18.6: 15 findings across 7 tests — text clipping on every screen, contrast on Inspect and Settings, and a hit-region failure on the Dashboard at accessibility text sizes. 2026-07-20T20:52:38+00:00 Christian Cleberg hello@cleberg.net 2026-07-20T20:47:24+00:00 urn:sha1:d9bd61cc745b3f1d6ff481fd300abf3216d07d48 DomainDig.storekit sat beside the DomainDig/ source folder, and Xcode treated the two as conflicting when creating the synced configuration — the resulting confirmation dialog offered to replace the folder. Naming the file SyncedProducts.storekit removes the collision entirely and matches the convention Xcode uses for synced configurations elsewhere. Updates the project file reference and the scheme's StoreKitConfigurationFileReference to match. 2026-07-20T20:52:38+00:00 Christian Cleberg hello@cleberg.net 2026-07-20T20:29:50+00:00 urn:sha1:f834c8a8223c40bd17b90be8f6ab7db843d0914c The StoreKitConfigurationFileReference added in #11 used one '../' too many, resolving outside the repository. Xcode resolves it relative to the .xcodeproj's xcshareddata directory, so two levels reaches the repo root. SweepActivityAttributes is now explicitly nonisolated. The app target sets SWIFT_DEFAULT_ACTOR_ISOLATION = MainActor while the widget target does not, so a type shared by both inferred a main-actor-isolated ActivityAttributes conformance that ActivityKit cannot use from its concurrent contexts. LocalAPIService's logger closures captured self strongly while their inner Tasks declared [weak self]. The weak capture is now on the outer closure and bound before the Task, so the concurrently-executing closure references an immutable strong local rather than the weak capture. 2026-07-20T20:07:40+00:00 Christian Cleberg hello@cleberg.net 2026-07-20T19:08:09+00:00 urn:sha1:b5d67f42b7f7e009eb60a80857a306414117969f Adds DomainDig.storekit mirroring the App Store Connect setup: one subscription group with Pro+ at level 1 and Pro at level 2, using the corrected product IDs. Wires it into the Run action so purchases resolve locally against StoreKit instead of the App Store, and disables the DOMAIN_DIG_FORCE_PRO_PLUS launch argument, which bypasses StoreKit entirely and would mask whether the purchase path works. Prices in the configuration are local-testing placeholders and do not need to match App Store Connect. 2026-04-24T20:04:54+00:00 Christian Cleberg hello@cleberg.net 2026-04-24T20:04:54+00:00 urn:sha1:3f0b7746b1bee4c2f626eb805e3fd13c5f2f5d22 * Stable domain intervals * Frequent changes decrease intervals * Duplicate states do not trigger alerts * Quiet hours suppress alerts * Quiet hours across midnight work correctly * Sensitivity levels alter behavior as expected