2043673 - Distinguish between "USers with no voices" and "USers where we couldn't populate voices"

Status: RESOLVED FIXED

(Core :: Privacy: Anti-Tracking, enhancement)

Description

[Tom Ritter [:tjr]]

Presently they both fall into the same values reported.

Comment 1

[Tom Ritter [:tjr]]

Attachment: Bug 2043673 - Distinguish timed-out voice enumeration from genuine empty r?timhuang

speechSynthesis.getVoices() populates asynchronously via voiceschanged.
populateVoiceList already races a 5s timeout against that population;
previously the timeout path resolved a fake empty list, making timed-out
submissions indistinguishable from devices with no TTS engine in the
characteristics ping (both reported voicesCount=0 / voicesSha1=sha1("")).

This patch makes the timeout path emit no voices_* fields at all, so
timed-out submissions appear as NULL voicesCount / voicesSha1 / etc. in
BigQuery and genuine empty enumerations remain voicesCount=0. The two
cohorts are now separable without overloading any field's semantic
meaning. populateMappableData's dispatch loop now skips any metric whose
populator returned undefined, which is purely additive for every other
populator.

Comment 2

[Pulsebot]

Pushed by [e-mail]:
https://github.com/mozilla-firefox/firefox/commit/5cd872b2df0a
https://hg.mozilla.org/integration/autoland/rev/33acd8d51258
Distinguish timed-out voice enumeration from genuine empty r=timhuang

Comment 3

[Norisz Fay [:noriszfay]]

https://hg.mozilla.org/mozilla-central/rev/33acd8d51258