GCC5: Phone restarts if you click on anything in the first page of welcome wizard

[Dave Morley]

STEPS:
1. Install the latest gcc5 test image
2. Get to the welcome wizard
3. Tap the language drop down
4. Reboots
5. Tap the Continue Button
6. Reboots

EXPECTED:
I expect the welcome wizard to work the same as it does on vivid

ACTUAL:
Any option you tap on as soon as the signal is received it reboots the phone.

VERSIONS:
Image 3

Logs to follow once I can strip them out of the device.

[kevin gunn (kgunn72)]

ok, after some discussion on irc
leaving this incomplete atm, b/c i can't tell what unity8 bins were used for this image.
in fact, unity8 bins built with gcc5 are stuck in transition.

[Steve Langasek (vorlon)]

This is an image built from wily-proposed, for purposes of validating before the gcc5 transition lands in wily.

[Gerry Boland (gerboland)]

The welcome wizard is part of unity8. unity8 usually loads a QPA plugin called "mirserver" from QtMir, so that is it a mir server.

This log suggests unity8 loaded the "ubuntumirclient" plugin, which is from QtUbuntu. Why it did this we need to figure out. The output of unity8.log would help, as would running unity8 with the QT_DEBUG_PLUGINS=1 env var set.

I'm guessing ubuntumirclient is loaded as a backup if mirclient fails to load.

Possible reasons:
1. mirclient QPA missing, meaning qtmir-{desktop,android} package not installed somehow
2. mirclient QPA plugin not loadable, dlload unhappy with it somehow

[kevin gunn (kgunn72)]

actually, i just ran here, crash file i see appears to be falling over at
ua_sensors_haptic_vibrate_once in platform api
i see that someone (kenvandine?) also had reason to suspect this and just did a no change rebuild....
it's not installed on my phone, so i'm gonna upgrade and see if we hit a new roadblock

[kevin gunn (kgunn72)]

OK, getting something different. Update didn't help, but i'm consistently getting this crash

[Alberto Aguirre (albaguirre)]

Stack trace:

#0 0xad3daf98 in deallocate (this=<optimized out>, __p=<optimized out>)
    at /usr/include/c++/5/ext/new_allocator.h:110
#1 deallocate (__a=..., __n=<optimized out>, __p=<optimized out>)
    at /usr/include/c++/5/bits/alloc_traits.h:386
#2 _M_destroy (__size=<optimized out>, this=<optimized out>)
    at /usr/include/c++/5/bits/basic_string.h:185
#3 _M_dispose (this=0xbe8fd794) at /usr/include/c++/5/bits/basic_string.h:180
#4 ~basic_string (this=0xbe8fd794, __in_chrg=<optimized out>)
    at /usr/include/c++/5/bits/basic_string.h:544
#5 interface_name () at /usr/include/core/dbus/traits/service.h:34
#6 invoke_method_synchronously<ubuntu::application::sensors::USensorD::Haptic::Vibrate, void, unsigned int> (this=<optimized out>)
    at /usr/include/core/dbus/impl/object.h:71
#7 ua_sensors_haptic_vibrate_once (sensor=<optimized out>,
    duration=<optimized out>)
    at /build/platform-api-qm_L3b/platform-api-3.0.0+15.10.20150730/src/ubuntu/application/common/application/sensors/service.cpp:93
#8 0xa8600e8e in QFeedbackHapticsEffect::setState(QFeedbackEffect::State) ()
   from /usr/lib/arm-linux-gnueabihf/libQt5Feedback.so.5
#9 0xa7d2deb6 in ?? ()
   from /usr/lib/arm-linux-gnueabihf/qt5/qml/QtFeedback/libdeclarative_feedback.so

[Alberto Aguirre (albaguirre)]

Also "test_ubuntu_haptic_device" which calls the same ua_sensors_haptic_vibrate_once api runs without crashing.

Some package versions:
http://pastebin.ubuntu.com/12003474/

[kevin gunn (kgunn72)]

the fix is now in proposed