[2026-10-16] eagle: work/projects/cpm-status-2026-10-15.md work/projects/cpm-web-extension-code-guide-2026-10-09.md
This commit is contained in:
1 parent
0ebd030fae
commit
c7da49d791
2 files changed
+38
-1
No files matched your search
@@ -602,3 +602,19 @@ The harness now posts and dequeues one neutral application-defined event after w
|
||||
### Signing verification detail
|
||||
|
||||
The latest signed build (`build-signed-event-loop.log`) invokes codesign with **Apple Development: Alexey Martemyanov (9SAT8X9S84)**, team HKE973VLUW, hardened runtime. `codesign -dvv` with normal keychain access confirms the Apple Development → WWDR → Apple Root CA chain. App debug dylib and `Integration Tests.xctest` independently pass `codesign --verify --strict`. `codesign --verify --deep --strict` on the entire Debug app fails in nested `DuckDuckGo VPN.app/Contents/Frameworks/Subscription.framework`: the seal lists four missing `Modules/Subscription.swiftmodule` artifacts (abi.json/swiftdoc/swiftmodule/Project/swiftsourceinfo). This is a separate nested Debug packaging limitation, **not an unsigned main app or test bundle**. No manual re-signing or production packaging changes were made for it.
|
||||
|
||||
|
||||
### Verification of cleanup/event-queue changes: still red
|
||||
|
||||
- `signed-pinned-final-control.xcresult`: posting/dequeuing a neutral event alone did **not** resolve the DEBUG window interruption. The first real P5 failure is again recorded in stdout, but no completed green test. The memory graph's currentEvent chain is a real reference; it was not established as the sole cause. Harness synchronous `Tab.setContent`, diagnostics collection and polling conditions are now also confined to autorelease pools, pending verification.
|
||||
- `signed-large-cache-cleanup.xcresult`: the 20k cache scenario again completed its body with healthy post-reload/readiness controls, but static store removal still failed after the bounded retry. Merely clearing data and dropping harness references was insufficient: the pre-existing Integration Tests process-pool swizzle outlives each test. Teardown now requests `_terminateNetworkProcess` on the owned store **after evidence capture and data clearing**, before UUID removal. This is cleanup, not a causal fault used to claim P3. The previous fresh UUID `0FB56422-72D5-4CCB-AC66-86F3C50DE165` is recorded explicitly for cleanup auditing; no user-store identifier is selected.
|
||||
|
||||
|
||||
WebKit cleanup contract checked against primary sources: [profiles API](https://webkit.org/blog/14423/building-profiles-with-new-webkit-api/) documents that removal can fail while an active view uses the profile; [WKWebsiteDataStore.mm](https://github.com/WebKit/WebKit/blob/main/Source/WebKit/UIProcess/API/Cocoa/WKWebsiteDataStore.mm#L1088) implements `_terminateNetworkProcess` by forwarding to its WebsiteDataStore. Current upstream source is reference material, not an assertion that this macOS release has the identical revision. The actual runtime selector is checked before test-only cleanup.
|
||||
|
||||
|
||||
### P5 app test scope narrowed explicitly, not a hidden green result
|
||||
|
||||
`signed-pinned-scoped-pools.xcresult` again interrupted after the first lost navigation. The harness printed `currentEventStillOwnsWindow=false`, so advancing the event queue demonstrably does **not** solve the remaining lifetime problem. That neutral-event workaround is removed. The earlier graph also showed the closed MainWindow directly retained in an autorelease pool while the async XCTest was running; the sole original autorelease call site was not established. Synchronous harness navigation/diagnostics/polling remain inside scoped pools as normal lifetime hygiene.
|
||||
|
||||
The app P5 method is now deliberately limited to **one** real failed navigation (`tab_in_context=false`, matching controller, `bg_view_creations=2`) after two healthy/idle cycles and actual shared-pin window closure, followed by healthy regular/new-tab controls. It no longer attempts the second four-second timeout that reaches the app's eight-second DEBUG window guard before XCTest can finish. The guard is not disabled, no controller/tab detach is fabricated, and no claim is made that app-hosted persistent P5/stuck passed. The separate source-linked SPM P5 tests retain the existing two-failure/stuck proof. A new app run is still required for the narrowed method.
|
||||
Reference in new issue
Block a user