diff options
| author | Robert Carr <racarr@google.com> | 2018-02-28 18:06:10 -0800 |
|---|---|---|
| committer | Robert Carr <racarr@google.com> | 2018-03-01 12:43:40 -0800 |
| commit | 99742734e3b67cc48941e37db982c43aa843596c (patch) | |
| tree | 3c3b01b5c52a4113a1ef2f9a9b8958f5495b0dbc /tools/aapt2/java/JavaClassGenerator_test.cpp | |
| parent | 824095856be2745050d5c45d91bd88492f2940f4 (diff) | |
Fix toast lifetime
In NotificationManagerService#cancelToast we have been calling
WindowManagerService#removeWindowToken with 'removeWindows'=true. This
is allowing for Surface destruction without any sort of synchronization
from the client. Before the call to removeWindowToken we are emitting
a one-way hide call to the Toast client. As a solution to the lifetime
issue we have the client callback to let us know it has processed the
hide call (and thus stopped the ViewRoot). On the server side we also instate
a timeout. This mirrors the app stop timeout. All codepaths I could find
leading to this sort of situation where a client is still rendering
in to a toast following the total duration expiring seem to indicate a hung
client UI thread.
Bug: 62536731
Bug: 70530552
Test: Manual. go/wm-smoke
Change-Id: I89643b3c3a9fa42423b498c1bd3a422a7959aaaf
Diffstat (limited to 'tools/aapt2/java/JavaClassGenerator_test.cpp')
0 files changed, 0 insertions, 0 deletions