diff --git a/docs/website/data/port_status_supplement.json b/docs/website/data/port_status_supplement.json index 094957269d3..d54f900c280 100644 --- a/docs/website/data/port_status_supplement.json +++ b/docs/website/data/port_status_supplement.json @@ -58,7 +58,7 @@ }, { "test": "GoogleWebMapScreenshotTest", - "reason": "The live Google Maps JavaScript map needs three things at once: a web view that composites into the screenshot, a billed Google Maps API key, and network access to Google's tile servers. The deterministic desktop and web pipelines are not given the key, the Mac Catalyst and browser captures do not composite the native web view, tvOS ships no WebKit, and the watch form factor has no map surface at all.", + "reason": "The live Google Maps JavaScript map needs three things at once: a web view that composites into the screenshot, a billed Google Maps API key, and network access to Google's tile servers. Every desktop port (macOS, Mac Catalyst, Linux and Windows) and the browser skip it before reading a key, because their captures do not composite the native web view; only the Android and iOS pipelines are given the key (no-api-key is still accepted from Linux and Windows reports written before they took the desktop path), tvOS ships no WebKit, and the watch form factor has no map surface at all.", "platform_support": "Every listed port supports BrowserComponent and the map API where the table says so; this skip is about one live, key-gated and network-dependent third-party map rendering inside a screenshot comparison, not about map support.", "verification": "The Android and iOS legs are given the API key and render the live Google map end to end, comparing the painted tiles against a stored golden. The offline vector maps that need no key are exercised on every port by the VectorMap and RealOsmVector tests. Where the key and a web view are present, a map that never finishes loading is recorded as this skip rather than as a failure: after a full wait budget and a retry on a warm web view, what the run has established is that the tiles could not be fetched, which is a property of the runner's network rather than of the port. A map that does load is compared against its golden as strictly as before, so a rendering change still fails on the pixels.", "reason_codes": [ @@ -76,7 +76,11 @@ "ports": [ "macos", "mac-catalyst", - "javascript" + "javascript", + "linux-x64", + "linux-arm64", + "windows-x64", + "windows-arm64" ] }, { diff --git a/scripts/hellocodenameone/conformance/test_port_status.py b/scripts/hellocodenameone/conformance/test_port_status.py index a23bf5ef0f5..a2681c2341b 100755 --- a/scripts/hellocodenameone/conformance/test_port_status.py +++ b/scripts/hellocodenameone/conformance/test_port_status.py @@ -618,7 +618,7 @@ def test_an_undocumented_skip_accounts_for_nothing(self): def test_a_skip_documented_for_another_port_accounts_for_nothing(self): # map-tiles-never-loaded is written about android and the two iOS renderers. The same - # reason arriving from Linux, where the errata expect no-api-key instead, is a port + # reason arriving from Linux, where the errata expect a desktop reason instead, is a port # behaving unexpectedly rather than a network nobody can reach. self.assertEqual( [],