There is a race condition in NetworkMonitor where native observers may be removed concurrently with a notification being dispatched, leading to a dangling pointer dereference (trying to dispatch an observer that was already removed and destroyed), and from there a crash with access violation.
By ensuring dispatching to native observers is done within the synchronization lock that guards additions/removals of native observers protects against this race condition.
Since native observers callbacks are posted to the networking thread in the C++ side anyway, there should be no risk of deadlock/starvation due to long-running observers.
Bug: webrtc:15837
Change-Id: Id2b788f102dbd25de76ceed434c4cd68aa9a569e
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/338643
Reviewed-by: Taylor Brandstetter <deadbeef@webrtc.org>
Commit-Queue: Harald Alvestrand <hta@webrtc.org>
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#42256}
94dfe1cc59