1052046 - geolocation/GPS on Flame no longer works

Status: RESOLVED WORKSFORME

(Core :: DOM: Geolocation, defect)

Description

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

[Blocking Requested - why for this release]:

The GPS on the flame phone used to be great; it would sync within a few seconds.

Now it doesn't give me a GPS fix at all.  I think this regressed within the past few weeks, although it could be a little bit longer.  I'm running master gecko/gaia on the v122 base image plus the new fonts.

I'm currently running:
gaia 2d2475b521351e200136e463358e6c8e91957702
gecko a6424bfa8f39 plus local patches as of Fri Aug 08 16:43:07 2014 -0700

Comment 1

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Reverting https://hg.mozilla.org/mozilla-central/rev/051fd57f4fef locally makes this work again.

Comment 2

[Chris Peterson [:cpeterson]]

Kan-Ru: do you have a Flame device you can use to reproduce this GPS problem?

Bug 1032063 was uplifted to the 1.4 release branch, so we may need to back out or fix it there, too.

Comment 3

[Kan-Ru Chen [:kanru] (UTC+9)]

I think on Flame we are using QC RIL and QC location so this change shouldn't affect Flame. I'm not sure if mozRIL works with Flame at all.

:dbaron could you list the /system/b2g/distribution/bundles directory?

% adb shell ls /system/b2g/distribution/bundles
b2g_time
libqc_b2g_location
libqc_b2g_ril

Comment 4

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

I see:

$ adb shell ls /system/b2g/distribution/bundles
b2g_time
libqc_b2g_location
libqc_b2g_ril

(I've only ever flashed Gecko, Gaia, and the new fonts on top of the v122 base image, so I believe I shuold still be using the QC RIL, etc.)

Comment 5

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

And today it works.  Maybe I just wasn't waiting long enough?

Comment 6

[:garvan]

It is very possible this bug is related to: https://bugzilla.mozilla.org/show_bug.cgi?id=1033274

Why? I can reliably repo that bug only when acquiring AGPS data. In David's case, I believe he was seeing that bug, then AGPS loaded fully, and it was masked. Note that bug 1033274 is not believed to be GPS code precisely, rather it is an event and sequencing bug, resulting in JS callback objects becoming unhooked.