GitLab

This defensive fix until MediaSession owners return from vacation.
Reverting is challenging given the long string of CLs.

GetMediaPlayerRemote(player_id) is crashing for failing to find the
remote with that ID. The stack shows the call coming from
MediaSessionController, which should have the same lifetime as its
associated mojo::Remote<media::mojom::MediaPlayer>. So far, I have not
found how those lifetimes are getting out of sync.

Ownership looks like this (read -> as "owns", MWCO as MediaWebContentsObserver):
MWCO -> MediaSessionControllersManager -> map<MediaPlayerId, MediaSessionController(s)>
MWCO -> map<MediaPlayerId, mojo::Remote<media::mojom::MediaPlayer>>

I've combed through all the places where MWCO removes from its map of
MediaPlayer remotes, and looked for a failure to call
MediaSessionControllersManger::OnEnd(Id) to remove the associated
MediaSessionController. There are 3 cases where we remove from the
map...
1. MediaWebContentsObserver::RenderFrameDeleted()
2. MediaWebContentsObserver::WebContentsDestroyed()
3. media_player_remotes_[player_id].set_disconnect_handler(base::BindOnce(...

The only case where I don't see a call to
MediaSessionControllersManager is #2. But that shouldn't matter since
MWCO's destructor is called shortly (synchronously) after
WebContentsDestroyed(). Its destructor will tear down the
MediaSessionControllersManager.

The defensive fix does not address the root lifetime issue. It is
probably the case that the MediaSessionControllers are being leaked.

Bug: 1161551, 1039252
Change-Id: I85337eae106d8b56b7d48fc9759702dc579855ca
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/2607579
Commit-Queue: Chrome Cunningham <chcunningham@chromium.org>
Commit-Queue: Frank Liberato <liberato@chromium.org>
Reviewed-by: Frank Liberato <liberato@chromium.org>
Reviewed-by: Daniel Cheng <dcheng@chromium.org>
Auto-Submit: Chrome Cunningham <chcunningham@chromium.org>
Cr-Commit-Position: refs/heads/master@{#839818}