832987 - With "don't keep activities", possible to put the browser in a bad state where no pages load

Status: RESOLVED FIXED

(Firefox for Android Graveyard :: General, defect)

Description

[Kartikaya Gupta (email:[e-mail])]

On a Galaxy Nexus, enable the developer options and turn on the "don't keep activities" option. Then start FF and immediately tap on the awesomebar to enter a URL. Enter some URL (e.g. people.mozilla.com) and hit ok/go to load it.

Expected results:
Page loads

Actual results:
Nothing loads, and the browser stays on about:home. Furthermore, any other attempts to load pages after this don't work either.

It seems like having the GeckoApp activity get killed in the middle of Gecko startup puts us into a bad state from which we can't recover.

Comment 1

[Kartikaya Gupta (email:[e-mail])]

It looks like this is happening because GeckoApp is destroyed when the awesomebar is entered, and so it never receives and processes the Gecko:Ready event. The sLaunchState remains in Launching and so events that are sent to gecko just keep getting tacked onto the pending events list. Yay for updating persistent state from a non-persistent listener/object!

Comment 2

[Kartikaya Gupta (email:[e-mail])]

Attachment: Patch

This moves the launch state stuff into GeckoThread, so that the two critical lines in the Gecko:Ready handler can run even if the GeckoApp instance has been destroyed at the time the Gecko:Ready message arrives.

Comment 3

[Kartikaya Gupta (email:[e-mail])]

Try build at https://tbpl.mozilla.org/?tree=Try&rev=e39a6fe411b8

Comment 4

[Chris Peterson [:cpeterson]]

Comment on attachment 705024 [details] [diff] [review]
Patch

Review of attachment 705024 [details] [diff] [review]:
-----------------------------------------------------------------

LGTM

Comment 5

[Kartikaya Gupta (email:[e-mail])]

Try build showed that I forgot to fix BaseTest.blockForGeckoReady's reflection code. Fixed that up and re-pushed to try at https://tbpl.mozilla.org/?tree=Try&rev=2289881bf39e - it looks green enough, so landed on m-i:

https://hg.mozilla.org/integration/mozilla-inbound/rev/a2cc396069a2

Comment 6

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/a2cc396069a2