зеркало из https://github.com/mozilla/gecko-dev.git
cb268d10f4
It looks like there could be a race condition where GetTerminationPromise() is called at the same time on two threads on an empty MozPromiseHolder, and one of the MozPromises created by MozPromiseHolder::Ensure if leaked (and its chained callbacks never called). Additionally, I don't think the ParentStatusProtected() > Running check is necessary because for ServiceWorkers, the global can't call close(), so the parent status is at least "Canceling", making the CloseWorkerOnMainThread call unnecessary (which just calls WorkerPrivate::Cancel). The parent process also will always send a termination operation as the last operation, so we can just rely on that to make sure the worker is fully terminated. Differential Revision: https://phabricator.services.mozilla.com/D56277 --HG-- extra : moz-landing-system : lando |
||
|---|---|---|
| .. | ||
| PRemoteWorker.ipdl | ||
| PRemoteWorkerController.ipdl | ||
| PRemoteWorkerService.ipdl | ||
| RemoteWorkerChild.cpp | ||
| RemoteWorkerChild.h | ||
| RemoteWorkerController.cpp | ||
| RemoteWorkerController.h | ||
| RemoteWorkerControllerChild.cpp | ||
| RemoteWorkerControllerChild.h | ||
| RemoteWorkerControllerParent.cpp | ||
| RemoteWorkerControllerParent.h | ||
| RemoteWorkerManager.cpp | ||
| RemoteWorkerManager.h | ||
| RemoteWorkerParent.cpp | ||
| RemoteWorkerParent.h | ||
| RemoteWorkerService.cpp | ||
| RemoteWorkerService.h | ||
| RemoteWorkerServiceChild.cpp | ||
| RemoteWorkerServiceChild.h | ||
| RemoteWorkerServiceParent.cpp | ||
| RemoteWorkerServiceParent.h | ||
| RemoteWorkerTypes.ipdlh | ||
| moz.build | ||