chore: version packages

Merged
GI
9d ago
+115-161

This PR was opened by the Changesets release GitHub action. When you're ready to do a release, you can merge this and the packages will be published to npm automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to main, this PR will be updated.

Releases

@e2e-dev/decision@0.2.4

Patch Changes

  • #1035 7e568f2 Thanks @rudycelekli! - Count the tokens and served model of a model call that returns while its step is being cancelled.

e2e@0.21.2

Patch Changes

  • #957 b1646e2 Thanks @zichen0116! - The agent's tap, double_tap, and right_click tools, in agent.act and the MCP server, take an optional modifiers list (Shift, Control, Alt, Meta, ControlOrMeta) on an engine that declares tapModifiers, like the test API's tap({ modifiers }). The held keys are recorded and replayed from the cache. Custom executors see the capability as ctx.target.tapModifiers and pass { modifiers } to ctx.actions.tap, doubleTap, and secondaryTap.

  • #1192 40bea9e Thanks @adc77! - Report the model behind cached actions and identify replays made with a different model.

  • #1029 7d067af Thanks @rudycelekli! - Align failure code-frame carets and clip source lines by terminal display width.

  • #1190 e4d16ec Thanks @okwasniewski! - A TypeScript file that another loader (such as tsx) left with an inline source map that is not JSON or not shaped like a source map now loads, with stack traces pointing at that loader's output, instead of failing with a TypeError.

  • #1153 bf3c8e8 Thanks @TheTuso! - copilot(): a model GitHub Copilot does not serve to the account (model_not_supported) now stops the run once with MODEL_UNAVAILABLE naming the model and the cause, instead of failing every test with MODEL_PROVIDER_FAILED.

  • #1178 2791786 Thanks @michalpobuta! - The getByTestId and SemanticNode.testId docs now say an engine may also match an id without its platform's namespacing. No type changes.

  • #1168 236611f Thanks @okwasniewski! - Correct two messages: a missing engine version no longer claims it keys the replay cache, and the output option's description names results/ instead of artifacts/.

  • #1187 82d1f3f Thanks @okwasniewski! - With more than one worker, e2e run starts the files that took longest in the previous run first, so a slow file no longer starts last and holds the run open after the other workers finish.

  • #1184 3272c16 Thanks @okwasniewski! - The e2e mcp server is listed in the official MCP Registry as io.github.tester-army/e2e.

  • #1027 7e967d9 Thanks @rudycelekli! - Keep multiline test titles on one row in the terminal live window so repainting erases the complete previous block.

  • #1190 e4d16ec Thanks @okwasniewski! - Subscription sign-in, token refresh, and e2e models fail with FLOW_FAILED naming the vendor when a successful response's body is not JSON or not the expected shape, instead of a TypeError or SyntaxError. e2e models skips a model entry it cannot read, and opencodeConsole() treats a malformed workspace config like one it could not fetch.

  • #1186 d64e4e5 Thanks @okwasniewski! - Subscription sign-in's code exchange and token refresh fail with FLOW_FAILED when the token endpoint answers null, instead of a TypeError. A refresh_token or id_token that is not a string is ignored.

  • #1168 236611f Thanks @okwasniewski! - The packaged docs now include the snippets/ their pages import, so the model provider list reads offline, and several pages are corrected against the current APIs.

  • #1182 11c0951 Thanks @TheTuso! - The markdown summary, the GitHub comment, and the trace page name the step whose error ended the test. A step failure the test caught and went on from no longer stands in as the failure's step and line.

  • #1189 e626053 Thanks @okwasniewski! - scrollUntilVisible no longer pauses 100 ms between swipes. On web, a swipe resolves once the scroll it caused is on screen, so the next read sees the scrolled page.

  • #1156 b750125 Thanks @bheemreddy-samsara! - After close and a successful trace write, session.report() returns the hosted session's Report.

  • #1113 af0437b Thanks @jayhemnani9910! - --shard now reports a skipped test in one shard only, the one that runs the nearest test before it (the first shard when none does). Every shard used to list it, so a skipped test was counted once per shard in the summaries, junit.xml, and report.json.

@e2e-dev/eas@0.3.2

Patch Changes

  • #1190 e4d16ec Thanks @okwasniewski! - EAS Simulators fails with EAS: unexpected response to <operation> when Expo's API answers without the fields a session needs, instead of a TypeError, and a daemon address answered as null reads as not ready yet. Limrun names the installed agent-device package.json when it declares no version or no agent-device bin, instead of reporting version undefined or not finding the CLI.

  • #897 2219744 Thanks @rudycelekli! - Reject invalid minute limits before starting EAS sessions, preserving zero and valid integer settings.

@e2e-dev/github@0.4.2

Patch Changes

  • #1038 aa9699a Thanks @rudycelekli! - Distinguish GitHub primary and secondary API rate limits from read-only token failures in reporter diagnostics.

@e2e-dev/lightpanda@0.0.1

Patch Changes

  • #1163 b41afa2 Thanks @okwasniewski! - New package: lightpanda(), a browser provider that starts a lightpanda serve per worker slot on a free port and stops it when the run ends.

@e2e-dev/limrun@0.2.1

Patch Changes

  • #1181 fc5c350 Thanks @okwasniewski! - Bump agent-device to 0.21.28.

  • #1190 e4d16ec Thanks @okwasniewski! - EAS Simulators fails with EAS: unexpected response to <operation> when Expo's API answers without the fields a session needs, instead of a TypeError, and a daemon address answered as null reads as not ready yet. Limrun names the installed agent-device package.json when it declares no version or no agent-device bin, instead of reporting version undefined or not finding the CLI.

@e2e-dev/mobile@0.12.3

Patch Changes

  • #1166 4fe6baf Thanks @okwasniewski! - Locators and expect now keep waiting while Android reports insufficient foreground app content (an app still loading after a cold start) instead of failing at once. A wait that runs out on an unreadable screen fails with that reason.

  • #1181 fc5c350 Thanks @okwasniewski! - Bump agent-device to 0.21.28.

  • #1178 2791786 Thanks @michalpobuta! - getByTestId on Android also matches a resource id without its <package>:id/ prefix when nothing matches the full id, so a .NET MAUI AutomationId or a native android:id matches the same string as on iOS. A negative check on such a name, like toHaveCount(0), can now see a node it used to miss, and two prefixed views that share the name on one screen are ambiguous.

  • #1183 cf5e122 Thanks @spsaucier! - An iOS screen's location takes its title from the navigation bar's identifier only when the bar shows that text, not from the bar's label or a class name left in the identifier, and an Android toolbar's resource id is no longer a title; iOS recordings that start or end on a pushed screen, and Android recordings whose path ends in a toolbar resource id, re-record once (--strict-cache runs fail with REPLAY_STALE until then).

@e2e-dev/testmu@0.0.1

Patch Changes

  • #772 88ed717 Thanks @amankansal-lt! - New @e2e-dev/testmu package: mobile({ device: testmu({ device, osVersion, app }) }) runs a mobile target on TestMu AI (formerly LambdaTest) Android emulators, iOS simulators, and real devices, one per worker slot, with LT_USERNAME and LT_ACCESS_KEY from the run's environment. An attempt that records video links TestMu AI's recording of its session.

@e2e-dev/web@0.15.2

Patch Changes

  • #1164 de7afd6 Thanks @TheTuso! - web({ connect }) is INVALID_CONFIG at config load when cdpEndpoint is missing or not a function, instead of failing every attempt with ENGINE_FAILURE: connection failed: resolve is not a function (with reconnectEndpoint) or worker init with browser connect failed: source.resolve is not a function (without it).

  • #1189 e626053 Thanks @okwasniewski! - scrollUntilVisible no longer pauses 100 ms between swipes. On web, a swipe resolves once the scroll it caused is on screen, so the next read sees the scrolled page.

  • #1188 52a0c7d Thanks @okwasniewski! - A swipe on a node the pointer already hovers, or one hit at its centre, wheels it without hovering first, so each scrollUntilVisible stride is faster.