App Domain .callServiceExtension .restart .log #212
Conversation
| Future<Map<String, dynamic>> _restart(Map<String, dynamic> args) async { | ||
| var appId = getStringArg(args, 'appId', required: true); | ||
| if (_appId != appId) throw ArgumentError.value(appId, 'appId', 'Not found'); | ||
| // TODO(grouma) - Figure out what fullRestart means in this context. |
There was a problem hiding this comment.
Ah, full restart was the old name of hot restart. So, fullRestart==false is hot reload, and fullRestart==true is hot restart.
There was a problem hiding this comment.
Thanks for the details. I updated the code to reflect our current support.
| // For now we will ignore. | ||
| // var fullRestart = getBoolArg(args, 'fullRestart') ?? false; | ||
| var pauseAfterRestart = getBoolArg(args, 'pause') ?? false; | ||
| if (pauseAfterRestart) { |
There was a problem hiding this comment.
You may not want to throw here - we'd have to investigate. The IDEs will be sending it in when you start the app in a debug launch (as opposed to a 'run', w/o breakpoints enabled).
If the IDE does pass in pauseAfterRestart, it will wait until it gets a PausePostRequest event, remove all breakpoints, re-set all breakpoints, then resume the isolate. It does this to ensure that the new scripts, representing libraries that had been reloaded, have the correct breakpoints set.
There was a problem hiding this comment.
Worth seeing what happens here first before doing a lot of work however.
There was a problem hiding this comment.
I can try this out Monday, but in VS Code we do always send pause when running in debug mode (to do what Devon describes above). The reason for this is that we need to re-send breakpoints after a hot restart (they got dropped) or hot reload (the code they were attached to may have been replaced with new code), and if we don't pause then we'll have race conditions (the code may have already run before we've sent the breakpoint).
| @@ -43,6 +48,13 @@ class AppDomain extends Domain { | |||
| sendEvent('app.started', { | |||
| 'appId': _appId, | |||
| }); | |||
| _vmService.onStdoutEvent.listen((log) { | |||
There was a problem hiding this comment.
nit: move this next to the streamListen?
| webdev.stdin.add(utf8.encode('$extensionCall\n')); | ||
| // The example app sets up a service extension for printing. | ||
| await expectLater( | ||
| webdev.stdout, emitsThrough(startsWith('[{"event":"app.log"'))); |
There was a problem hiding this comment.
This could easily pass for a different reason than the extension being called successfully (any log happening), I would check a bit more about the event if you can (check specifically for the expected print log line)
Towards #191