Closed Bug 936324 Opened 11 years ago Closed 11 years ago

[Browser] Keyboard no longer appears on gecko inbound + gaia master builds

Categories

(Firefox OS Graveyard :: General, defect)

Product:
Firefox OS Graveyard
Component:
General
Platform:
ARM
Gonk (Firefox OS)
Type:
defect
Priority:
Not set
Severity:
blocker

Tracking

(blocking-b2g:1.3+, firefox28 fixed)

Status:
RESOLVED FIXED
Project Flags:
blocking-b2g 1.3+
Tracking Flags:
Tracking Status
firefox28 --- fixed

People

(Reporter: nbp, Assigned: timdream)

References

Details

(Keywords: regression) 

I am building Unagi images out of inbound and gaia master.

STR:
 - Open the browser app.
 - Tap on the url bar.

Expected:
 - The keyboard should appear with the blinking cursor in the url bar.

Seen:
 - The blinking cursor appear, but not the keyboard.

This breaks Marionette harness for running JavaScript benchmarks in the browser application.  If needed I can give a copy of the last image used to flash the device.

No error message appears in adb, except a failure of Marionette caused by the missing keyboard.

Gecko regression range inferred from AWFY:
http://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=471db633b20c&tochange=a751d2c77ec4
blocking-b2g: --- → 1.3?
This issue seems to be isolated to the browser app.
Also note that AWFY is using repo sync, which means that the version of gaia tested are not the one visible in the regression range above, but they should be more recent than the first push of the previous regression range.
Component: Gaia → Gaia::Browser
Summary: Keyboard no longer appears on gecko inbound + gaia master builds → [Browser] Keyboard no longer appears on gecko inbound + gaia master builds
Flags: needinfo?(timdream)
I will try to flash a phone with b2g-inbound and bisect Gaia, if that's what cause the issue.
Flags: needinfo?(timdream)
I also met this from Settings App.
I can confirm that the current m-c + gaia master works, so I think is a Gecko problem which I will try to bisect.
Assignee: nobody → timdream
bisecting...
Component: Gaia::Browser → Gaia::Keyboard
No, this is a Gecko issue.
Component: Gaia::Keyboard → General
Here is my bisect result -- Keyboard failed on all runs I tried so it's a bit strange. The susceptible offending commit is bug 895082.

$ git bisect log
git bisect start
# bad: [4984e7db61c3df8b7804d06770c463560bd18748] Bug 880031 - Make animated transforms force all above content to be in a separate ThebesLayer. r=roc
git bisect bad 4984e7db61c3df8b7804d06770c463560bd18748
# good: [b7effd61f438a18ac3facd5da4856233d075b74f] Merge fx-team to m-c
git bisect good b7effd61f438a18ac3facd5da4856233d075b74f
# bad: [c1aa7dd3e7fc3bf3b9b92479907fbf6a7bbfeb1a] Bug 935297 - Part 5: Implement StreamToSink on CG. r=Bas,jrmuizel
git bisect bad c1aa7dd3e7fc3bf3b9b92479907fbf6a7bbfeb1a
# bad: [375dc70580ad9daab1f3d3aa67d6527c8b0312a0] Merge m-c to inbound
git bisect bad 375dc70580ad9daab1f3d3aa67d6527c8b0312a0
# bad: [205e94be742f509db3bb78e9ef45972069431842] Bug 932112:  Use the non-main-thread FindClass implementation r=blassey
git bisect bad 205e94be742f509db3bb78e9ef45972069431842
# bad: [18be1240fdd85d5cda0b4b3e9adefea48edc5d37] Bug 936031 - Test case for bug. r=abr
git bisect bad 18be1240fdd85d5cda0b4b3e9adefea48edc5d37
# bad: [7d09aa8e5c3645e52903d14879ca55e504033148] bug 931792 - only init sCanaryOutputFD once r=bsmedberg
git bisect bad 7d09aa8e5c3645e52903d14879ca55e504033148
# bad: [08cd5e8af4a2eae610602e8e8952b4f6231ecabf] bug 895082 - recursively cache children in DocAccessible::ProcessContentInserted() r=surkov
git bisect bad 08cd5e8af4a2eae610602e8e8952b4f6231ecabf
Assignee: timdream → nobody
OK, I don't think I bisect Gecko correctly.
This last good commit should be

# good: [b61c9756dbe9ec01429ba066276e2ea2a641ff4a] bug 934716 - generateCRMFRequest: accept only the first repeated keygen parameter r=cviecco
git bisect good b61c9756dbe9ec01429ba066276e2ea2a641ff4a
Assignee: nobody → timdream
Note that we are talking about hashes on git@github.com:mozilla/integration-gecko-dev.git
Judging by the current |git bisect visualize|, I suspect bug 932151 is at fault. :yxl, did you test that patch for in-process apps (i.e. the browser?)
Flags: needinfo?(xyuan)
Confirmed.

560a2ee4625f5ed04ad0ef061df3c41d0af0b371 is the first bad commit
commit 560a2ee4625f5ed04ad0ef061df3c41d0af0b371
Author: Yuan Xulei <xyuan@mozilla.com>
Date:   Thu Nov 7 09:02:25 2013 -0500

    Bug 932151 - Include forms.js in both desktop and b2g build. r=fabrice
    
    --HG--
    rename : b2g/chrome/content/forms.js => dom/inputmethod/forms.js

:040000 040000 5b24c2f29d0aa840ecf314e246f24b39881420b2 54d13287ee6b59caf5fa1653db4185f1db4da750 M      b2g
:040000 040000 d7b9a73ecf0d8af2b230f2399da9a48422f9419a dc8e9c920ac600a0950562297fe6d0a1dc97a6e4 M      dom


git bisect start
# bad: [4984e7db61c3df8b7804d06770c463560bd18748] Bug 880031 - Make animated transforms force all above content to be in a separate ThebesLayer. r=roc
git bisect bad 4984e7db61c3df8b7804d06770c463560bd18748
# bad: [c1aa7dd3e7fc3bf3b9b92479907fbf6a7bbfeb1a] Bug 935297 - Part 5: Implement StreamToSink on CG. r=Bas,jrmuizel
git bisect bad c1aa7dd3e7fc3bf3b9b92479907fbf6a7bbfeb1a
# bad: [375dc70580ad9daab1f3d3aa67d6527c8b0312a0] Merge m-c to inbound
git bisect bad 375dc70580ad9daab1f3d3aa67d6527c8b0312a0
# bad: [205e94be742f509db3bb78e9ef45972069431842] Bug 932112:  Use the non-main-thread FindClass implementation r=blassey
git bisect bad 205e94be742f509db3bb78e9ef45972069431842
# bad: [18be1240fdd85d5cda0b4b3e9adefea48edc5d37] Bug 936031 - Test case for bug. r=abr
git bisect bad 18be1240fdd85d5cda0b4b3e9adefea48edc5d37
# bad: [7d09aa8e5c3645e52903d14879ca55e504033148] bug 931792 - only init sCanaryOutputFD once r=bsmedberg
git bisect bad 7d09aa8e5c3645e52903d14879ca55e504033148
# bad: [08cd5e8af4a2eae610602e8e8952b4f6231ecabf] bug 895082 - recursively cache children in DocAccessible::ProcessContentInserted() r=surkov
git bisect bad 08cd5e8af4a2eae610602e8e8952b4f6231ecabf
# bad: [1b903b6054664b94356468f3aea01027f5d6c488] Backed out changeset ec8ac6112088 (bug 933885) for frequent xpcshell test_singlebytes.js timeouts.
git bisect bad 1b903b6054664b94356468f3aea01027f5d6c488
# good: [b61c9756dbe9ec01429ba066276e2ea2a641ff4a] bug 934716 - generateCRMFRequest: accept only the first repeated keygen parameter r=cviecco
git bisect good b61c9756dbe9ec01429ba066276e2ea2a641ff4a
# bad: [e9612c5af33d1ee68fe7c833f4d990b3494d5847] Bumping gaia.json for 4 gaia-central revision(s) a=gaia-bump
git bisect bad e9612c5af33d1ee68fe7c833f4d990b3494d5847
# good: [0fa1b8bdd250a426e775be49c17f8fb8ed5f5979] Bug 921637 - Move activities so gaia can consume them in any environment. r=sicking
git bisect good 0fa1b8bdd250a426e775be49c17f8fb8ed5f5979
# bad: [560a2ee4625f5ed04ad0ef061df3c41d0af0b371] Bug 932151 - Include forms.js in both desktop and b2g build. r=fabrice
git bisect bad 560a2ee4625f5ed04ad0ef061df3c41d0af0b371
Blocks: 932151
Fixed by backout of bug 932151.
Status: NEW → RESOLVED
Closed: 11 years ago
Resolution: --- → FIXED
blocking-b2g: 1.3? → ---
Flags: needinfo?(xyuan)
(In reply to Tim Guan-tin Chien [:timdream] (MoCo-TPE) (please ni?) from comment #12)
> Judging by the current |git bisect visualize|, I suspect bug 932151 is at
> fault. :yxl, did you test that patch for in-process apps (i.e. the browser?)
No.

Does the patch only break the browser app's keyboard or all apps' keyboard?
The patch doesn't change the behaviour of the keyboard or related API. It just
moved one file. If that break the the browser's keyboard, it should also break other
app's keyboard.
For the sake of tracking blocking regressions, I'm going to + this.

Context - QA reviews blocking regressions for a release to help identify automation gaps. So if it makes sense to block even if the bug was backed out, we should then still plan to + the bug.
blocking-b2g: --- → 1.3+
Keywords: regression
This has hit Travis now so people are going to want a backout asap.
(In reply to Yuan Xulei [:yxl] from comment #15)
> (In reply to Tim Guan-tin Chien [:timdream] (MoCo-TPE) (please ni?) from
> comment #12)
[snip]
> Does the patch only break the browser app's keyboard or all apps' keyboard?
> The patch doesn't change the behaviour of the keyboard or related API. It
> just
> moved one file. If that break the the browser's keyboard, it should also
> break other
> app's keyboard.

From what I can see using the latest build, only the browser app is affected relating to keyboard. All the other apps I have used (Contacts, Email, etc) I can use the keyboard.

Gaia   f7b408fe7e40a6e783e3f29c5cea5ec5ba5349ae
SourceStamp 70f21fad60a4
BuildID 20131108040203
Version 28.0a1
(In reply to Zac C (:zac) from comment #18)
> This has hit Travis now so people are going to want a backout asap.

The backout that happened above didn't work then, did it? Travis is still red?
(In reply to Jason Smith [:jsmith] from comment #20)
> (In reply to Zac C (:zac) from comment #18)
> > This has hit Travis now so people are going to want a backout asap.
> 
> The backout that happened above didn't work then, did it? Travis is still
> red?

I might have jumped the gun a bit. Saw some similar failures before but it seems OK now.
(In reply to Ed Morley [:edmorley UTC+1] from comment #14)
> Fixed by backout of bug 932151.

AreWeFastYet harness is working again, since the merge of b2g-inbound in mozilla-inbound.
You need to log in before you can comment on or make changes to this bug.