1150315 - Successively calling speechSynthesis speak() and cancel() brings queue into bad state

Status: RESOLVED FIXED

(Core :: DOM: Core & HTML, defect)

Description

[Eitan Isaacson [:eeejay]]

The first call to speak() puts one utterance in the queue and calls AdvanceQueue(). AdvanceQueue() in turn peeks at the first utterance in the queue (it only gets popped when the utterance ends), and assigns mCurrentTask with a new task with that utterance.

After calling cancel() the speech task cancels the utterance and waits for an end-of-stream before doing anything further. In the meantime speak() is called again, and mCurrentTask is reassigned.

When the end-of-stream arrives, and OnEnd() is called and the task argument does not match the current task member since it was swapped out by speak().

Comment 1

[Eitan Isaacson [:eeejay]]

Attachment: Keep SpeechSynthesis state in sync after successive cancel/speak calls.

Comment 2

[Eitan Isaacson [:eeejay]]

This happens (and the test fails) when the speech synthesis is in an OOP child window. When it is all in one process, it happens synchronously, so the state doesn't get messed up.

Comment 3

[Eitan Isaacson [:eeejay]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/e96b90efdeec

Comment 4

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/e96b90efdeec