mir_stress suite causes mir_demo_server to crash

[Thomi Richards]

Running the mir_stress suite causes the mir_demo_server binary to crash.

I'm building lp:mir from trunk, and running mir_demo_server and the mir_stress binaries. See attached stack trace for details.

[Robert Ancell (robert-ancell)]

I'm not able to reproduce. I've done basically:

$ bzr branch lp:mir
$ cd mir
$ ./native-compile.sh
$ sudo ./build-linux-x86/bin/mir_demo_server

And from another VT:
$ sudo ./build-linux-x86/bin/mir_stress

[Thomi Richards (thomir-deactivatedaccount)]

It seems you need to run mir_stress a second time before it will crash the server.

[Robert Ancell (robert-ancell)]

I had to run mir_stress a number of times but it does now reproduce for me.

[Robert Ancell (robert-ancell)]

This seems to be related to the threading in the frontend, still investigating...

[Robert Carr (robertcarr)]

Found the cause (short story: session_manager::close_session can throw and SessionMediator::disconnect is unequipped to deal with it. SessionManager::close_session is throwing due to a few seperate races though), think I've found a solution for it all. Will have a proposal tomorrow :)

[Olli Ries (ories)]

this (mir stress tests) will become part of the release criteria - and thus will require some attention.
Robert A/C, was there any progress?

[Francis Ginther (fginther)]

The qa team is working to enable the mir stress test as a default part of the mir-ci job to give feedback on future merge proposals. Eventually this will also be enabled for merging to trunk (i.e. mir-autolanding). We can coordinate timing to allow the mir team to resolve this.

[Daniel van Vugt (vanvugt)]

Seems no longer in progress. Also, I just found a machine which can reproduce the crash...

terminate called after throwing an instance of 'boost::exception_detail::clone_impl<boost::exception_detail::error_info_injector<std::logic_error> >'
  what(): Invalid Session

#0 0x00007fdd59921f77 in __GI_raise (sig=sig@entry=6)
    at ../nptl/sysdeps/unix/sysv/linux/raise.c:56
#1 0x00007fdd599255e8 in __GI_abort () at abort.c:90
#2 0x00007fdd59f286e5 in __gnu_cxx::__verbose_terminate_handler() ()
   from /usr/lib/x86_64-linux-gnu/libstdc++.so.6
#3 0x00007fdd59f26856 in ?? () from /usr/lib/x86_64-linux-gnu/libstdc++.so.6
#4 0x00007fdd59f25919 in ?? () from /usr/lib/x86_64-linux-gnu/libstdc++.so.6
#5 0x00007fdd59f264ca in __gxx_personality_v0 ()
   from /usr/lib/x86_64-linux-gnu/libstdc++.so.6
#6 0x00007fdd59cc17f3 in ?? () from /lib/x86_64-linux-gnu/libgcc_s.so.1
#7 0x00007fdd59cc1d27 in _Unwind_Resume ()
   from /lib/x86_64-linux-gnu/libgcc_s.so.1
#8 0x00007fdd5ad2d537 in mir::shell::SessionManager::close_session (
    this=0x1aa41d8, session=...)
    at /home/dan/bzr/mir/trunk/src/server/shell/session_manager.cpp:130
#9 0x00007fdd5acd583b in mir::frontend::SessionMediator::~SessionMediator (
    this=0x7fdd2c003408, __in_chrg=<optimised out>)
    at /home/dan/bzr/mir/trunk/src/server/frontend/session_mediator.cpp:74
#10 0x00007fdd5acae7e7 in __gnu_cxx::new_allocator<mir::frontend::SessionMediator>::destroy<mir::frontend::SessionMediator> (this=0x7fdd2c003400,
    __p=0x7fdd2c003408) at /usr/include/c++/4.8/ext/new_allocator.h:124
#11 0x00007fdd5acacfdb in std::allocator_traits<std::allocator<mir::frontend::SessionMediator> >::_S_destroy<mir::frontend::SessionMediator> (__a=...,
    __p=0x7fdd2c003408) at /usr/include/c++/4.8/bits/alloc_traits.h:281
#12 0x00007fdd5acaba0e in std::allocator_traits<std::allocator<mir::frontend::SessionMediator> >::destroy<mir::frontend::SessionMediator> (__a=...,
    __p=0x7fdd2c003408) at /usr/include/c++/4.8/bits/alloc_traits.h:405
#13 0x00007fdd5aca9775 in std::_Sp_counted_ptr_inplace<mir::frontend::SessionMediator, std::allocator<mir::frontend::SessionMediator>, (__gnu_cxx::_Lock_policy)2>::_M_dispose (this=0x7fdd2c0033f0)
    at /usr/include/c++/4.8/bits/shared_ptr_base.h:407
#14 0x0000000000452b0e in std::_Sp_counted_base<(__gnu_cxx::_Lock_policy)2>::_M_release (this=0x7fdd2c0033f0)
    at /usr/include/c++/4.8/bits/shared_ptr_base.h:144
#15 0x0000000000452441 in std::__shared_count<(__gnu_cxx::_Lock_policy)2>::~__shared_count (this=0x7fdd2c003558, __in_chrg=<optimised out>)
    at /usr/include/c++/4.8/bits/shared_ptr_base.h:553
#16 0x00007fdd5acdadde in std::__shared_ptr<mir::protobuf::DisplayServer, (__gnu_cxx::_Lock_policy)2>::~__shared_ptr (this=0x7fdd2c003550,
    __in_chrg=<optimised out>)
    at /usr/include/c++/4.8/bits/shared_ptr_base.h:810
#17 0x00007fdd5acdadf8 in std::shared_ptr<mir::protobuf::DisplayServer>::~shared_ptr (this=0x7fdd2c003550, __in_chrg=<optimised out>)
    at /usr/include/c++/4.8/bits/shared_ptr.h:93
...

[Robert Carr (robertcarr)]

I think there is a good chance that this is fixed now . I'm unable to reproduce it so far. I think the changes in focus control flow in client-focus-notifications fix the main issue that I was seeing (an invalid surface exception is thrown out of focus and handled up in the frontend, preventing the session mediator from clearing it's .session after disconnect and before ~SessionMediator which then tries to destroy the session again but it's gone).

[Robert Carr (robertcarr)]

IT was too good to be true, disregard that.

[Daniel van Vugt (vanvugt)]

The bug also manifests as:

terminate called after throwing an instance of 'boost::exception_detail::clone_impl<boost::exception_detail::error_info_injector<std::logic_error> >'
  what(): Requesting handle for an unregistered channel

If you run mir_stress with some input on another client.

[PS Jenkins bot (ps-jenkins)]

Fix committed into lp:mir at revision None, scheduled for release in mir, milestone 0.0.10

[Launchpad Janitor (janitor)]

This bug was fixed in the package mir - 0.0.10+13.10.20130827.1-0ubuntu1

---------------
mir (0.0.10+13.10.20130827.1-0ubuntu1) saucy; urgency=low

  [ Alan Griffiths ]
  * ipc: add a protocol version to the wire protocol so that we can bump
    it in future.
  * graphics::nested: Handling of output configuration changes.
  * graphics.nested: Hookup NestedDisplay to display change
    notifications.

  [ Daniel van Vugt ]
  * Introducing mir_demo_client_progressbar. It's pretty boring;
    designed to simulate key repeat scrolling in a terminal, as an aid
    for tracking down bug 1216472. . (LP: #1216472)

  [ Eleni Maria Stea ]
  * changed the GBMBufferAllocator constructor and class to use the
    gbm_device instead of the GBMPlatform to remove the dependency from
    the mg::Platform interface - this way we can use the
    GBMBufferAllocator with the NativeGBMPlatform (nested mir).

  [ Ubuntu daily release ]
  * Automatic snapshot from revision 1025
 -- Ubuntu daily release <email address hidden> Tue, 27 Aug 2013 18:04:47 +0000