indicator-location crashing during default, sdk and click_image_tests tests on smoketesting

Unfortunately, apport-retrace thinks that all three of these crash reports have corrupt cores and don't provide useful backtraces. All of them show that /something/ called abort(), but the retrace fails before we get to the caller:

SourcePackage: indicator-location
Stacktrace:
 #0 0xb6b258e6 in ?? () from /lib/arm-linux-gnueabihf/libc.so.6
 No symbol table info available.
 #1 0xb6b3405e in raise () from /lib/arm-linux-gnueabihf/libc.so.6
 No symbol table info available.
 #2 0xb6b34d4e in abort () from /lib/arm-linux-gnueabihf/libc.so.6
 No symbol table info available.
 #3 0xb6c622b4 in __gnu_cxx::__verbose_terminate_handler() () from /usr/lib/arm-linux-gnueabihf/libstdc++.so.6
 No symbol table info available.
 #4 0xb6c60af0 in ?? () from /usr/lib/arm-linux-gnueabihf/libstdc++.so.6
 No symbol table info available.
 Backtrace stopped: previous frame identical to this frame (corrupt stack?)

Same for all three traces.

FWIW there's no direct call to abort(), assert(), g_error() in the indicator-location codebase, so the abort() call is coming from somewhere else in the stack.

So, thostr_ noticed something I didn't :-P by looking at the indicator-location logs:

terminate called after throwing an instance of 'std::runtime_error'
  what(): org.freedesktop.DBus.Error.ServiceUnknown: The name com.ubuntu.location.Service was not provided by any .service files
terminate called after throwing an instance of 'std::runtime_error'
  what(): org.freedesktop.DBus.Error.ServiceUnknown: The name com.ubuntu.location.Service was not provided by any .service files
terminate called after throwing an instance of 'std::runtime_error'
  what(): org.freedesktop.DBus.Error.ServiceUnknown: The name com.ubuntu.location.Service was not provided by any .service files
terminate called after throwing an instance of 'std::runtime_error'
  what(): org.freedesktop.DBus.Error.ServiceUnknown: The name com.ubuntu.location.Service was not provided by any .service files

It looks like we brute forcing this by adding location service to indicator-datetime's debian/control file would work, but I'm not sure that's the right fix: from indicator-datetime's point of view, location-service isn't used directly, but rather is hidden behind platform-api's public API.

So it seems cleaner for com.ubuntu.location.Service to be in platform-api's dependencies, and only /indirectly/ in indicator-location's dependencies via platform-api.

Adding Thomas to this ticket for a second opinion...

I think the real problem is not the service not being available, but the platform API not handling the error case gracefully. Bumped Priority to "High", will introduce a more sensible error handling strategy.

It's very likely this problem on errors.u.c:

https://errors.ubuntu.com/problem/7fbb29cd16437f0b85c68abcc20d4d0b6cea0c13 and bug 1344047 is a duplicate

confirmed bug 1344047 is a dupe with charles.

This bug was fixed in the package platform-api - 2.2.0+14.10.20140730-0ubuntu1

---------------
platform-api (2.2.0+14.10.20140730-0ubuntu1) utopic; urgency=low

  [ thomas-voss ]
  * Check for null in functions accessing a session object. Make sure
    that exceptions cannot propagate for controller interactions. (LP:
    #1338610)
 -- Ubuntu daily release <email address hidden> Wed, 30 Jul 2014 07:26:34 +0000

Note, not tagging for our qa-daily-testing/rtm14 as user interaction isn't affected.