Skip to content

YC (S26) | Open Computer History | Continuously record your company computer work, map your workflows, help you find work worth automating, and power your agents' context

fix(ci): unblock E2E (register new spec in coverage-map + stop ort system-linking on Windows)

Louis Beaumont4adada83720e
diff --git a/.github/workflows/e2e-test.yml b/.github/workflows/e2e-test.yml
index f7c538bd6..6456d7fd3 100644
--- a/.github/workflows/e2e-test.yml
+++ b/.github/workflows/e2e-test.yml
@@ -146,11 +146,13 @@ jobs:
if (Test-Path $ortDir) { Remove-Item $ortDir -Recurse -Force }
Copy-Item -Path $tauriDir -Destination $ortDir -Recurse -Force
- - name: Set ORT env vars
- shell: bash
- run: |
- echo "ORT_STRATEGY=system" >> $GITHUB_ENV
- echo "ORT_LIB_LOCATION=${{ github.workspace }}/ort-system-lib" >> $GITHUB_ENV
+ # No ORT_STRATEGY/ORT_LIB_LOCATION: link ONNX Runtime via ort's
+ # download-binaries (pyke prebuilt), same as the green release-app and the
+ # ci.yml/windows-integration/release-cli workflows. Setting ORT_LIB_LOCATION
+ # forces ort-sys into system static-linking, which needs onnxruntime_common.lib
+ # + the static _deps tree the dynamic MS prebuilt doesn't ship -> "ort-sys
+ # could not link" (the e2e Windows Build failure). The MS zip downloaded above
+ # is still staged below for the runtime DLL.
- name: Cache Next.js build
uses: actions/cache@v4
@@ -172,9 +174,9 @@ jobs:
projectPath: "./apps/screenpipe-app-tauri"
tauriScript: bunx tauri -v
- # ORT_STRATEGY=system links against onnxruntime.lib but does NOT copy
- # onnxruntime.dll next to the test binary the way the default
- # download-binaries+copy-dylibs path would. Without a copy here the
+ # ort links via download-binaries but the audio crate's Windows ort
+ # features omit copy-dylibs, so onnxruntime.dll is NOT copied next to the
+ # test binary automatically. Without a copy here the
# Windows DLL loader fails to find onnxruntime.dll at runtime and the
# screenpipe-server thread panics with "Failed to initialize ORT API"
# on the very first ort::api() call. Tauri's bundle step does this for
diff --git a/apps/screenpipe-app-tauri/e2e/COVERAGE.md b/apps/screenpipe-app-tauri/e2e/COVERAGE.md
index aa1fafd87..e0b5dc5ff 100644
--- a/apps/screenpipe-app-tauri/e2e/COVERAGE.md
+++ b/apps/screenpipe-app-tauri/e2e/COVERAGE.md
@@ -6,9 +6,9 @@ and layer declared in the manifest, weighted by confidence and criticality.
- Manifest: `e2e/coverage-map.json`
- Specs directory: `e2e/specs`
-- Mapped specs: 50
-- Declared test blocks: 175
-- Weighted coverage points: 137.3
+- Mapped specs: 51
+- Declared test blocks: 176
+- Weighted coverage points: 138.0
Confidence weights: strong=1.0, partial=0.7, conditional=0.4, smoke=0.3.
Criticality weights: high=1.0, medium=0.7, low=0.4.
@@ -19,9 +19,9 @@ can execute more runtime cases than this number shows.
| Platform | Specs | Declared tests | Weighted points | Layers | Features | Critical score |
| --- | --- | --- | --- | --- | --- | --- |
-| windows | 43 | 164 | 133.9 | 15 | 51 | 92% |
-| macos | 47 | 141 | 109.9 | 15 | 51 | 89% |
-| linux | 37 | 129 | 106.3 | 13 | 48 | 86% |
+| windows | 44 | 165 | 134.6 | 15 | 52 | 92% |
+| macos | 48 | 142 | 110.6 | 15 | 52 | 89% |
+| linux | 38 | 130 | 107.0 | 13 | 49 | 86% |
## Runtime Results
@@ -34,7 +34,7 @@ pass/fail/skip counts.
| Layer | windows | macos | linux |
| --- | --- | --- | --- |
| audio-device | 2 specs / 26 tests / 19.4 pts | 1 specs / 1 tests / 0.3 pts | - |
-| billing | 1 specs / 1 tests / 1.0 pts | 1 specs / 1 tests / 1.0 pts | 1 specs / 1 tests / 1.0 pts |
+| billing | 2 specs / 2 tests / 1.7 pts | 2 specs / 2 tests / 1.7 pts | 2 specs / 2 tests / 1.7 pts |
| capture-ocr | 2 specs / 13 tests / 5.2 pts | 2 specs / 3 tests / 1.2 pts | 1 specs / 2 tests / 0.8 pts |
| chat-ai | 8 specs / 8 tests / 4.9 pts | 10 specs / 11 tests / 5.8 pts | 8 specs / 8 tests / 4.9 pts |
| local-api | 12 specs / 88 tests / 74.0 pts | 11 specs / 63 tests / 55.0 pts | 10 specs / 62 tests / 54.6 pts |
@@ -43,8 +43,8 @@ pass/fail/skip counts.
| os-integration | 3 specs / 16 tests / 15.1 pts | 3 specs / 3 tests / 0.9 pts | - |
| performance | 2 specs / 43 tests / 43.0 pts | 4 specs / 33 tests / 29.5 pts | 1 specs / 28 tests / 28.0 pts |
| pipes | 2 specs / 11 tests / 11.0 pts | 2 specs / 11 tests / 11.0 pts | 2 specs / 11 tests / 11.0 pts |
-| real-ui-e2e | 24 specs / 88 tests / 72.2 pts | 25 specs / 75 tests / 61.7 pts | 21 specs / 69 tests / 59.8 pts |
-| settings | 9 specs / 26 tests / 24.2 pts | 9 specs / 19 tests / 16.5 pts | 8 specs / 18 tests / 16.2 pts |
+| real-ui-e2e | 25 specs / 89 tests / 72.9 pts | 26 specs / 76 tests / 62.4 pts | 22 specs / 70 tests / 60.5 pts |
+| settings | 10 specs / 27 tests / 24.9 pts | 10 specs / 20 tests / 17.2 pts | 9 specs / 19 tests / 16.9 pts |
| storage-privacy | 6 specs / 20 tests / 19.1 pts | 5 specs / 12 tests / 11.1 pts | 4 specs / 12 tests / 11.1 pts |
| tauri-command | 8 specs / 17 tests / 10.3 pts | 9 specs / 19 tests / 10.8 pts | 8 specs / 17 tests / 10.3 pts |
| window-lifecycle | 16 specs / 60 tests / 51.2 pts | 16 specs / 41 tests / 29.6 pts | 12 specs / 36 tests / 28.1 pts |
@@ -136,6 +136,7 @@ pass/fail/skip counts.
| windows-core-recording.spec.ts | windows | capture-ocr, local-api, audio-device, real-ui-e2e | capture-ocr, local-api-auth, local-api-search, audio-device-health, timeline | high | conditional | mixed | 11 | Windows recording-enabled lane; hosted runners can skip frame-dependent OCR assertions. |
| windows-system-integration.spec.ts | windows | os-integration, local-api, audio-device, window-lifecycle, performance | app-launch, local-api-auth, audio-device-health, window-lifecycle, os-process-health, webview-stability | high | strong | mixed | 15 | Windows display, WebView2, loopback, process, Defender, audio, focus, and crash-report checks. |
| windows-user-journey.spec.ts | windows | real-ui-e2e, settings, notifications, storage-privacy, window-lifecycle | home-search, timeline, settings-recording, meeting-notes, shortcut-reminder, notifications, storage-retention, settings-privacy-api-auth | high | strong | real-user-flow | 8 | Windows-first real UX journey across search, timeline, settings, meetings, notifications, storage, and privacy. |
+| zz-account-stale-subscription.spec.ts | windows, macos, linux | real-ui-e2e, settings, billing | account-card, billing-gate | medium | strong | real-user-flow | 1 | Account 'active' plan card is gated on the session token (token AND cloud_subscribed) like the header, so a tokenless-but-subscribed stale shell (post-#3943 secret-store token desync) renders 'not logged in' with the active card gone. Mocks /api/user across all windows and clears set_cloud_token to reproduce the desync deterministically. |
| zz-app-entitlement-gate.spec.ts | windows, macos, linux | settings, billing, real-ui-e2e | app-entitlement-gate, billing-gate | high | strong | real-user-flow | 1 | Production billing gate blocks an unentitled session behind the paywall and restores access when the forced-gate flag is cleared. |
| zz-logout-resurrect.spec.ts | windows, macos, linux | real-ui-e2e, settings | account-logout, auth-session | high | strong | synthetic | 1 | Logout must not be resurrected by an in-flight loadUser. Logs in via a synthetic deep-link (?api_key=) with a mocked /api/user fetch, makes the fetch slow, fires it, then clicks logout while it is pending and asserts the slow response cannot re-write the user (the 'logout needs two clicks' bug). Covers the auth-generation guard in use-settings.tsx. |
| zz-owned-browser-background-nav.spec.ts | windows, macos | os-integration, window-lifecycle | owned-browser, window-lifecycle | low | smoke | command | 0 | Owned browser background navigation visibility. |
diff --git a/apps/screenpipe-app-tauri/e2e/coverage-map.json b/apps/screenpipe-app-tauri/e2e/coverage-map.json
index b25112993..a38aeb6f4 100644
--- a/apps/screenpipe-app-tauri/e2e/coverage-map.json
+++ b/apps/screenpipe-app-tauri/e2e/coverage-map.json
@@ -611,6 +611,16 @@
"confidence": "strong",
"ux": "synthetic",
"notes": "Logout must not be resurrected by an in-flight loadUser. Logs in via a synthetic deep-link (?api_key=) with a mocked /api/user fetch, makes the fetch slow, fires it, then clicks logout while it is pending and asserts the slow response cannot re-write the user (the 'logout needs two clicks' bug). Covers the auth-generation guard in use-settings.tsx."
+ },
+ {
+ "spec": "zz-account-stale-subscription.spec.ts",
+ "platforms": ["windows", "macos", "linux"],
+ "layers": ["real-ui-e2e", "settings", "billing"],
+ "features": ["account-card", "billing-gate"],
+ "criticality": "medium",
+ "confidence": "strong",
+ "ux": "real-user-flow",
+ "notes": "Account 'active' plan card is gated on the session token (token AND cloud_subscribed) like the header, so a tokenless-but-subscribed stale shell (post-#3943 secret-store token desync) renders 'not logged in' with the active card gone. Mocks /api/user across all windows and clears set_cloud_token to reproduce the desync deterministically."
}
]
}