| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
All three product targets (app, widget, share extension) now build under
SWIFT_VERSION = 6.0 with zero errors and zero warnings. The UITests
target stays on 5.0: XCTestCase's nonisolated setUp/init overrides
conflict with the target's MainActor default isolation under 6, and test
tooling is not shipping code.
The original seven diagnostics, plus the layers Swift 6 mode surfaced
once those cleared:
- SMTPChannel is an actor. It was implicitly MainActor while running its
receive loop on a background queue, so parsedLines/lineWaiters/
receiveBuffer were declared main-actor-protected and mutated off it —
concurrent mutation while resuming a CheckedContinuation can
double-resume, which traps. The actor serialises all state; Network
callbacks hop in via Task. The start() continuation also gains an
OSAllocatedUnfairLock resume-once guard: the state handler can fire
.ready and later .failed, and resuming twice was a pre-existing trap of
the same family.
- CachedLookupResult is nonisolated (a value pair built inside actor
LookupRuntime cannot have a MainActor-bound memberwise init) with
conditional Sendable — opting out of MainActor isolation also opted out
of the implicit Sendable that globally-isolated types get.
- PortScanService.printableBanner is nonisolated: a pure transformation
called from the connection's queue.
- SweepActivityController stores the activity's Sendable id instead of
the non-Sendable Activity, re-resolving via Activity.activities inside
each fire-and-forget task, so nothing non-Sendable crosses isolation.
- App Intents' static title/description/openAppWhenRun become lets
(get-only protocol requirements; static var is shared mutable global
state), and the summary helpers are @MainActor to match the model
properties they read and the perform() implementations that call them.
- ExternalDataService's ISO8601DateFormatter is nonisolated(unsafe),
citing Apple's documented thread-safety, rather than risking a parser
behaviour change by switching APIs with no test coverage.
- TaskMetricsDelegate.metrics is nonisolated(unsafe): written on the
session's delegate queue, read only after the request completes, and
URLSession guarantees didFinishCollecting precedes task completion.
- The share extension extracts the host via async/withCheckedContinuation
instead of sending a non-Sendable completion into loadItem's @Sendable
handler; Task inherits the view controller's MainActor so the manual
DispatchQueue.main hop goes too.
Validated: clean Swift 6 build of all product targets, and the full
enforced 11-test audit suite green on the floor runtime — Swift 6's
runtime isolation checks ran the app through every screen without a
trap.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
- New DomainReputationResult model (status: clean/listed/unknown, listed
sources, checked-at) and a `reputation(domain:)` method on ExternalDataService,
mirroring the existing pluggable-URL enrichment pattern (ownership history,
DNS history, extended subdomains, pricing). With no endpoint configured
(the default; DomainDig ships no bundled third-party reputation dependency)
it resolves to unavailable rather than "clean".
- New .reputation FeatureCapability/DataCapability, gated Pro+ like domainPricing.
- Threaded reputation/reputationError through LookupSnapshot and HistoryEntry
(backward-compatible decode) so results persist with history entries.
- Auto-fetched in performLookup alongside pricing; surfaced as a "Reputation"
info row, folded into DomainInsightEngine's risk score/factors and top-level
insights (a listed domain raises risk score and adds a factor/insight), and
exported in text, CSV, and JSON report output.
- Reputation-driven risk changes ride the existing change-severity pipeline, so
a listed status flip is visible to monitoring the same way any other risk
delta is, without bespoke monitoring wiring.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
historical context
and inferred enrichment.
- add derived intelligence fields for provider fingerprinting, classification,
ownership transitions, hosting transitions, subdomain history, risk signals,
and inferred timeline events
- expand DNS history beyond A/NS snapshots to retain A, AAAA, MX, NS, TXT, and
CNAME change state
- persist enriched intelligence in snapshots and history entries so analysis is
local-first and incremental
- add a dedicated Data+ Intelligence panel to current and historical domain
detail views
- surface intelligence events in timeline rows and include Data+ changes in diff
output
- preserve non-blocking inspection behavior by keeping enrichment additive to
the main lookup path
This makes Pro+ materially deeper for investigative workflows by improving
historical ownership visibility, infrastructure context, hosting change
detection, subdomain intelligence, and explainable risk signals.
|
|
|
- introduce Data+ tier for advanced data features
- add ownership history and DNS history
- expand subdomain discovery with external sources
- add domain pricing insights
- implement local usage/credit system
- integrate external data service layer with rate limiting
- extend export and CLI for Data+ features
|