race condition, using uninitialized Scene::jointMsgs containing NaN's

[osrf-migration]

Original report (archived issue) by John Hsu (Bitbucket: hsu, GitHub: hsu).

---

starting pr2 simulation sometimes leads to this segfault:

(gdb) /home/hsu/projects/groovy/simulator_gazebo/gazebo/scripts/gui: line 2: 28902 Aborted                 (core dumped) `rospack find gazebo`/gazebo/bin/gzclient -g `rospack find gazebo`/lib/libgazebo_ros_paths_plugin.so
[gazebo_gui-3] process has died [pid 28899, exit code 134, cmd /home/hsu/projects/groovy/simulator_gazebo/gazebo/scripts/gui __name:=gazebo_gui __log:=/home/hsu/.ros/log/dc93b228-513e-11e2-944f-90e2ba191409/gazebo_gui-3.log].
log file: /home/hsu/.ros/log/dc93b228-513e-11e2-944f-90e2ba191409/gazebo_gui-3*.log
bt
#0  0x00007ffff4f6e425 in __GI_raise (sig=<optimized out>) at ../nptl/sysdeps/unix/sysv/linux/raise.c:64
#1  0x00007ffff4f71b8b in __GI_abort () at abort.c:91
#2  0x00007ffff4f670ee in __assert_fail_base (fmt=<optimized out>, 
    assertion=0x7fffee9a63f0 "!pos.isNaN() && \"Invalid vector supplied as parameter\"", 
    file=0x7fffee9a62b0 "/tmp/buildd/ros-groovy-visualization-common-1.8.4/debian/ros-groovy-visualization-common/opt/ros/groovy/stacks/visualization_common/ogre/build/ogre_src_v1-7-3/OgreMain/src/OgreNode.cpp", line=<optimized out>, function=<optimized out>)
    at assert.c:94
#3  0x00007ffff4f67192 in __GI___assert_fail (
    assertion=0x7fffee9a63f0 "!pos.isNaN() && \"Invalid vector supplied as parameter\"", 
    file=0x7fffee9a62b0 "/tmp/buildd/ros-groovy-visualization-common-1.8.4/debian/ros-groovy-visualization-common/opt/ros/groovy/stacks/visualization_common/ogre/build/ogre_src_v1-7-3/OgreMain/src/OgreNode.cpp", line=413, 
    function=0x7fffee9a67e0 "virtual void Ogre::Node::setPosition(const Ogre::Vector3&)") at assert.c:103
#4  0x00007fffee7feb86 in setPosition (this=<optimized out>, pos=...)
    at /tmp/buildd/ros-groovy-visualization-common-1.8.4/debian/ros-groovy-visualization-common/opt/ros/groovy/stacks/visualization_common/ogre/build/ogre_src_v1-7-3/OgreMain/src/OgreNode.cpp:413
#5  Ogre::Node::setPosition (this=<optimized out>, pos=...)
    at /tmp/buildd/ros-groovy-visualization-common-1.8.4/debian/ros-groovy-visualization-common/opt/ros/groovy/stacks/visualization_common/ogre/build/ogre_src_v1-7-3/OgreMain/src/OgreNode.cpp:411
#6  0x00007fffee7fd972 in Ogre::Node::_setDerivedPosition (this=0x7fff48121f30, pos=...)
    at /tmp/buildd/ros-groovy-visualization-common-1.8.4/debian/ros-groovy-visualization-common/opt/ros/groovy/stacks/visualization_common/ogre/build/ogre_src_v1-7-3/OgreMain/src/OgreNode.cpp:548
#7  0x00007ffff29e4c86 in gazebo::rendering::Visual::SetWorldPosition (this=0xcbb2830, _pos=...)
    at /home/hsu/projects/groovy/simulator_gazebo/gazebo/build/gazebo-hg/gazebo/rendering/Visual.cc:1263
#8  0x00007ffff29645b9 in gazebo::rendering::JointVisual::Load (this=0xcbb2830, _msg=...)
    at /home/hsu/projects/groovy/simulator_gazebo/gazebo/build/gazebo-hg/gazebo/rendering/JointVisual.cc:55
#9  0x00007ffff29aa5fe in gazebo::rendering::Scene::ProcessJointMsg (this=0x16d8bb0, _msg=...)
    at /home/hsu/projects/groovy/simulator_gazebo/gazebo/build/gazebo-hg/gazebo/rendering/Scene.cc:1656
#10 0x00007ffff29a8be5 in gazebo::rendering::Scene::PreRender (this=0x16d8bb0)
    at /home/hsu/projects/groovy/simulator_gazebo/gazebo/build/gazebo-hg/gazebo/rendering/Scene.cc:1501
#11 0x00007ffff29ce118 in boost::_mfi::mf0<void, gazebo::rendering::Scene>::operator() (this=0x2945fe8, p=0x16d8bb0)
    at /usr/include/boost/bind/mem_fn_template.hpp:49
#12 0x00007ffff29ccd2e in boost::_bi::list1<boost::_bi::value<gazebo::rendering::Scene*> >::operator()<boost::_mfi::mf0<void, gazebo::rendering::Scene>, boost::_bi::list0> (this=0x2945ff8, f=..., a=...) at /usr/include/boost/bind/bind.hpp:253
#13 0x00007ffff29cb999 in boost::_bi::bind_t<void, boost::_mfi::mf0<void, gazebo::rendering::Scene>, boost::_bi::list1<boost::_bi::value<gazebo::rendering::Scene*> > >::operator() (this=0x2945fe8) at /usr/include/boost/bind/bind_template.hpp:20
#14 0x00007ffff29c9d8e in boost::detail::function::void_function_obj_invoker0<boost::_bi::bind_t<void, boost::_mfi::mf0<void, gazebo::rendering::Scene>, boost::_bi::list1<boost::_bi::value<gazebo::rendering::Scene*> > >, void>::invoke (function_obj_ptr=...)
    at /usr/include/boost/function/function_template.hpp:153
#15 0x00007ffff74242d4 in boost::function0<void>::operator() (this=0x2945fe0)
    at /usr/include/boost/function/function_template.hpp:1013
#16 0x00007ffff7427f0b in gazebo::event::EventT<void ()>::Signal() (this=0x7ffff7dd89e0)
    at /home/hsu/projects/groovy/simulator_gazebo/gazebo/build/gazebo-hg/gazebo/common/Event.hh:126
#17 0x00007ffff74258ce in gazebo::event::EventT<void ()>::operator()() (this=0x7ffff7dd89e0)
    at /home/hsu/projects/groovy/simulator_gazebo/gazebo/build/gazebo-hg/gazebo/common/Event.hh:119
#18 0x00007ffff6c65e94 in gazebo::sensors::SensorManager::Update (this=0x7ffff733f5a0, _force=false)
    at /home/hsu/projects/groovy/simulator_gazebo/gazebo/build/gazebo-hg/gazebo/sensors/SensorManager.cc:87
#19 0x00007ffff6c5ac2d in gazebo::sensors::run_once (_force=false)
    at /home/hsu/projects/groovy/simulator_gazebo/gazebo/build/gazebo-hg/gazebo/sensors/Sensors.cc:88
#20 0x00000000004697b1 in gazebo::Server::Run (this=0x6e96f0)
    at /home/hsu/projects/groovy/simulator_gazebo/gazebo/build/gazebo-hg/gazebo/Server.cc:371
#21 0x000000000046513f in main (argc=9, argv=0x7fffffffdcf8)
    at /home/hsu/projects/groovy/simulator_gazebo/gazebo/build/gazebo-hg/gazebo/server_main.cc:32
(gdb)

[osrf-migration]

Original comment by John Hsu (Bitbucket: hsu, GitHub: hsu).

---

it seems the very first jointMsg has corrputed position values, printing them out before usage yields:

Error [JointVisual.cc:52] -0.310343 -0.067658 3.83382e+174
Error [Visual.cc:1263] -0.310343 -0.067658 3.83382e+174

but if I put a break point on

  jointVis->Load(_msg);

inside of Scene::ProcessJointMsg(...), and step through the first couple of calls manually, then sometimes the position values are ok and simulation runs without segfault.

[osrf-migration]

Original comment by John Hsu (Bitbucket: hsu, GitHub: hsu).

---

Update

Turns out after tacking the ill formed values upstream, it was a typo on [parser_urdf.cc]https://github.com/osrf/gazebo/blob/12a2e98af84fb366d6e6e1e5bbc51fe2e630b198/gazebo/sdf/interface/parser_urdf.cc#L121)

Fixed in the branch.

The problem is because during the link loading loop, the call to link->load(linkElem), which calls Link::LoadCollision, which calls Entity::SetRelativePose, which ends up using potentially uninitialized parentEntity's worldPose.

[osrf-migration]

Original comment by John Hsu (Bitbucket: hsu, GitHub: hsu).

---

pull request #210 initializes Entity::worldPose during construction to avoid NaN's.

[osrf-migration]

Original comment by John Hsu (Bitbucket: hsu, GitHub: hsu).

---

• changed state from "new" to "resolved"

fixed in the branch

[osrf-migration]

Original comment by Nate Koenig (Bitbucket: Nathan Koenig).

---

• set version to "all"

[osrf-migration]

Original comment by Nate Koenig (Bitbucket: Nathan Koenig).

---

• changed state from "resolved" to "closed"