fenix
fenix copied to clipboard
[Bug]: Profiles with more than one first_run_date in the first session ping
Steps to reproduce
Querying the number of first_run_date per client_id in the first session ping returns more than one value.
WITH first_runs AS
(
SELECT
client_info.client_id, COUNT(1) AS dates
FROM
`moz-fx-data-shared-prod.org_mozilla_firefox_stable.first_session_v1`
WHERE
DATE(submission_timestamp) >= '2021-01-01'
AND DATE(submission_timestamp) = SAFE.PARSE_DATE('%F', SUBSTR(client_info.first_run_date, 1, 10))
AND ping_info.seq = 0
GROUP BY 1
)
SELECT
COUNT(client_id) clients,
COUNT(IF(dates>1, client_id, NULL)) AS clients_with_many_first_run_date
FROM first_runs
WHERE dates>1
Expected behaviour
Each client_id has a single first_run_date in moz-fx-data-shared-prod.org_mozilla_firefox_stable.first_session_v1
and a unique seq.
The code will only send the ping if it hasn't been sent yet.
Actual behaviour
There are 226 profiles with more than one first_session date and same seq between 2021 and 2022. (see query above)
Device name
NA
Android version
NA
Firefox release type
Firefox
Firefox version
NA
Device logs
NA
Additional information
This is related to a Glean first session ping.
┆Issue is synchronized with this Jira Task
I expect this is people that override our blocking of Android backup. This requires root.
Seems plausible, as the proportions we're seeing aren't inconsistent with such a niche population. Is there a way we could confirm by detecting their override, @kbrosnan ?
We don't have data on if the user is rooted.
Is it possible to instrument? I might dust off my Javalike language skills and implement such instrumentation if it'd give us confidence. But I don't know Android well enough to know if things like this will get us what we need.
Yes it is possible. Opened the above issue to look into collecting that data. It would be a few months before this hits release users.
Moved to bugzilla: https://bugzilla.mozilla.org/show_bug.cgi?id=1809326
Change performed by the Move to Bugzilla add-on.