+++ This bug was initially created as a clone of Bug #1023729 +++ Need to identify which global modifyLayout() used and move it's logic to layout_manager.js.

The good news is keyboardName is always the current layout but modifyLayout() takes the following globals. We need an API to get the following information to the new function. * currentInputType * currentInputMode * isGreekSMS() -> (currentInputMode === '-moz-sms' && keyboardName === 'el') * layoutPage (this is exposed to IMEngine with setLayoutPage()) * navigator.mozInputMethod.mgmt.supportsSwitching()

layoutPage has 3 values const LAYOUT_PAGE_DEFAULT = 'Default'; const LAYOUT_PAGE_SYMBOLS_I = 'Symbols_1'; const LAYOUT_PAGE_SYMBOLS_II = 'Symbols_2'; IMEngine only reference the first one.

Comment on attachment 8440363 [details] mozilla-b2g:master PR#20532 WIP https://travis-ci.org/mozilla-b2g/gaia/builds/27600292

It's likely I did something wrong because tests failed on Travis-CI. I don't know why Gaia-Try did run my push.

https://travis-ci.org/mozilla-b2g/gaia/builds/27607564 https://tbpl.mozilla.org/?tree=Gaia-Try&rev=0af02f6eac47c3be136ad10797c5b5bdc2f06a55

Comment on attachment 8440363 [details] mozilla-b2g:master PR#20532 Looks good to me. r+ Thanks.

master: https://github.com/mozilla-b2g/gaia/commit/a6c97a8924e9649052e7e33878ae4d28441975fd

Reverted from master in https://github.com/mozilla-b2g/gaia/commit/32ffcf9b8e01b350ba26673e6cbd1f61572b3265 under suspicion of causing the failures in bug 1027604

I don't think this patch cause the issue as TBPL comment keeps piling up since the revert. Backing out this patch will block bug 1025520 and onward. Please reland ASAP if we know this is false alarm.

I'm trying to reland this patch and fix the related issues. Will try to have several runs of Gu (Gaia ui tests) on TBPL and see the results. https://tbpl.mozilla.org/?tree=Gaia-Try&rev=886fca8ccec4e7aea8d176b4f17f66d16cfc4f94

Comment on attachment 8447049 [details] Reland patch It seems that b2g won't crash in Gaia ui test after running it more than 30 times on TBPL, https://tbpl.mozilla.org/?tree=Gaia-Try&rev=886fca8ccec4e7aea8d176b4f17f66d16cfc4f94 -- Tim, could you please help review this change? Thank you.

Comment on attachment 8447049 [details] Reland patch Great! Thanks for posting the tests. I would like the new pull request not by revert the revert but cherry-picking my individual commits because I want to preserve the right history. I can do that. I am also not entirely sure we will not hit bug 1027604 because I don't know if there is any difference between two. Let me create a pull request on that then.

New pull request created. https://tbpl.mozilla.org/?tree=Gaia-Try&rev=fae74583f5ab68c6ccd3b65bd84956e75d60c695

I also pushed to the TryServer so we can sure we are running the test on the same environment. https://tbpl.mozilla.org/?tree=Try&rev=55943ddc28aa

Wes, I don't see any crashes with the same summary like bug 1027604 in my try push. Could you help me to assess whether or not the patch is consider safe to land? Maybe bug 1027604 is fixed somewhere already...

This patch looks good to land to me, all of those failures are either known existing failures or infrastructure related.

Thanks! Let's try this one more time... master: https://github.com/mozilla-b2g/gaia/commit/b4a1d0df97c3d5bafc7a794cec368c9588d6b6b1