Allow retrying on "Operation timed out"

[d4rky-pl]

Use case

When opening a deep link to the app that still didn't finish fully loading (e.g. because CI is slow), it may sometimes/often fail with the following error:

com.github.michaelbull.result.UnwrapException: kotlin.Unit java.lang.IllegalStateException: An error was encountered processing the command (domain=NSPOSIXErrorDomain, code=60):01:23
Simulator device failed to open exp+helpling-customer://expo-development-client/?url=http%3A%2F%2F192.168.64.2%3A8081.01:23
Operation timed out01:23
Underlying error (domain=NSPOSIXErrorDomain, code=60):01:23
	The operation couldn’t be completed. Operation timed out01:23
	Operation timed out01:23
 01:23
	at maestro.drivers.IOSDriver$openLink$1.invoke(IOSDriver.kt:589)01:23
	at maestro.drivers.IOSDriver$openLink$1.invoke(IOSDriver.kt:435)01:23
	at maestro.utils.Metrics.measured(Metrics.kt:48)01:23
	at maestro.drivers.IOSDriver.openLink(IOSDriver.kt:435)01:23
	at maestro.Maestro.openLink(Maestro.kt:489)01:23
	at maestro.orchestra.Orchestra.openLinkCommand(Orchestra.kt:878)01:23
	at maestro.orchestra.Orchestra.executeCommand(Orchestra.kt:308)01:23
	at maestro.orchestra.Orchestra.executeSubflowCommands(Orchestra.kt:772)01:23
	at maestro.orchestra.Orchestra.runSubFlow(Orchestra.kt:826)01:23
	at maestro.orchestra.Orchestra.runFlowCommand(Orchestra.kt:671)01:23
	at maestro.orchestra.Orchestra.executeCommand(Orchestra.kt:316)01:23
	at maestro.orchestra.Orchestra.executeCommands(Orchestra.kt:212)01:23
	at maestro.orchestra.Orchestra.runFlow(Orchestra.kt:143)01:23
	at maestro.cli.runner.MaestroCommandRunner.runCommands(MaestroCommandRunner.kt:189)01:23
	at maestro.cli.runner.TestRunner$runSingle$result$1.invoke(TestRunner.kt:71)01:23
	at maestro.cli.runner.TestRunner$runSingle$result$1.invoke(TestRunner.kt:62)01:23
	at maestro.cli.runner.TestRunner.runCatching(TestRunner.kt:179)01:23
	at maestro.cli.runner.TestRunner.runSingle(TestRunner.kt:62)01:23
	at maestro.cli.runner.TestRunner.runSingle$default(TestRunner.kt:42)01:23
	at maestro.cli.command.RecordCommand$call$1.invoke(RecordCommand.kt:136)01:23
	at maestro.cli.command.RecordCommand$call$1.invoke(RecordCommand.kt:114)01:23
	at maestro.cli.session.MaestroSessionManager.newSession(MaestroSessionManager.kt:126)01:23
	at maestro.cli.session.MaestroSessionManager.newSession$default(MaestroSessionManager.kt:63)01:23
	at maestro.cli.command.RecordCommand.call(RecordCommand.kt:114)01:23
	at maestro.cli.command.RecordCommand.call(RecordCommand.kt:40)01:23
	at picocli.CommandLine.executeUserObject(CommandLine.java:1953)01:23
	at picocli.CommandLine.access$1300(CommandLine.java:145)01:23
	at picocli.CommandLine$RunLast.executeUserObjectOfLastSubcommandWithSameParent(CommandLine.java:2358)01:23
	at picocli.CommandLine$RunLast.handle(CommandLine.java:2352)01:23
	at picocli.CommandLine$RunLast.handle(CommandLine.java:2314)01:23
	at picocli.CommandLine$AbstractParseResultHandler.execute(CommandLine.java:2179)01:23
	at picocli.CommandLine$RunLast.execute(CommandLine.java:2316)01:23
	at maestro.cli.DisableAnsiMixin$Companion.executionStrategy(DisableAnsiMixin.kt:22)01:23
	at picocli.CommandLine.execute(CommandLine.java:2078)01:23
	at maestro.cli.AppKt.main(App.kt:154)

Proposal

Add an optional retry mechanism for those timeouts for when it's clear that the operation may take longer because the app is not ready yet

Anything else?

No response

[linear[bot]]

MAE-181 Allow retrying on "Operation timed out"

[Fishbowler]

What's the timeout here? Is it that the OS couldn't process the open link instruction because the Expo app wasn't fully metro'd up yet?

[Fishbowler]

Would love a way to reproduce this

[Fishbowler]

@d4rky-pl Any thoughts on how we'd reproduce this / test if we'd fixed it?

[d4rky-pl]

@Fishbowler hello and sorry for no reply, I kind of gave up on Maestro due to #1609 so I stopped checking notifications from this repository.

The problem was consistently happening on a resource starved CI machine (Semaphore CI). I'm not sure how to prepare a reliable repro but it was happening with an app using expo dev build and opening a deep link that tried to load the manifest from specific URL while the expo start in dev mode was running in the background.

[Fishbowler]

I suspect that running a Simulator and a JVM and doing some e2e tests between them might simply be unstable or impossible in a resource starved environment...