Segmentation fault on spawning model with collada mesh generated by assimp on ubuntu 18.04

[osrf-migration]

Original report (archived issue) by Naoki Hiraoka (Bitbucket: Naoki-Hiraoka).

---

related to ros/collada_urdf#34

When I spawned an urdf model with collada meshes into gazebo9 (ubuntu18.04), I got Segmentation fault (core dumped).

I generated the collada meshes with assimp v4.1.0.
I did not get this Segmentation fault on ubuntu 16.04 (gazebo7, assimp v3.2).

Thread 39 "gzserver" received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7fff1ffff700 (LWP 23344)]
0x00007ffff6a0aa76 in std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >::find_first_not_of(char, unsigned long) const ()
   from /usr/lib/x86_64-linux-gnu/libstdc++.so.6
(gdb) Segmentation fault (core dumped)
[gazebo_gui-3] process has died [pid 23197, exit code 139, cmd /opt/ros/melodic/lib/gazebo_ros/gzclient __name:=gazebo_gui __log:=/home/hiraoka/.ros/log/d975cdd0-31f4-11ea-99d9-5c514f34f247/gazebo_gui-3.log].
log file: /home/hiraoka/.ros/log/d975cdd0-31f4-11ea-99d9-5c514f34f247/gazebo_gui-3*.log

(gdb) bt
#0  0x00007ffff6a0aa76 in std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >::find_first_not_of(char, unsigned long) const ()
    at /usr/lib/x86_64-linux-gnu/libstdc++.so.6
#1  0x00007ffff6ec6acf in  ()
    at /usr/lib/x86_64-linux-gnu/libgazebo_common.so.9
#2  0x00007ffff6ed9a63 in gazebo::common::ColladaLoader::LoadPolylist(TiXmlElement*, ignition::math::v4::Matrix4<double> const&, gazebo::common::Mesh*) ()
    at /usr/lib/x86_64-linux-gnu/libgazebo_common.so.9
#3  0x00007ffff6edc859 in gazebo::common::ColladaLoader::LoadGeometry(TiXmlElement*, ignition::math::v4::Matrix4<double> const&, gazebo::common::Mesh*) ()
    at /usr/lib/x86_64-linux-gnu/libgazebo_common.so.9
#4  0x00007ffff6ede830 in gazebo::common::ColladaLoader::LoadNode(TiXmlElement*, gazebo::common::Mesh*, ignition::math::v4::Matrix4<double> const&) ()
    at /usr/lib/x86_64-linux-gnu/libgazebo_common.so.9
#5  0x00007ffff6edef89 in gazebo::common::ColladaLoader::LoadScene(gazebo::common::Mesh*) () at /usr/lib/x86_64-linux-gnu/libgazebo_common.so.9
#6  0x00007ffff6edf4b0 in gazebo::common::ColladaLoader::Load(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) ()
    at /usr/lib/x86_64-linux-gnu/libgazebo_common.so.9
#7  0x00007ffff6efe89a in gazebo::common::MeshManager::Load(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) ()
    at /usr/lib/x86_64-linux-gnu/libgazebo_common.so.9
#8  0x00007ffff5abee8b in gazebo::physics::MeshShape::Init() ()
    at /usr/lib/x86_64-linux-gnu/libgazebo_physics.so.9
#9  0x00007ffff59d54c4 in gazebo::physics::ODEMeshShape::Init() ()
    at /usr/lib/x86_64-linux-gnu/libgazebo_physics.so.9
#10 0x00007ffff5a65762 in gazebo::physics::Collision::Init() ()
    at /usr/lib/x86_64-linux-gnu/libgazebo_physics.so.9
#11 0x00007ffff5aac27e in gazebo::physics::Link::Init() ()
    at /usr/lib/x86_64-linux-gnu/libgazebo_physics.so.9
#12 0x00007ffff59d30ae in gazebo::physics::ODELink::Init() ()
    at /usr/lib/x86_64-linux-gnu/libgazebo_physics.so.9
#13 0x00007ffff5ac9a9e in gazebo::physics::Model::Init() ()
    at /usr/lib/x86_64-linux-gnu/libgazebo_physics.so.9
#14 0x00007ffff5b13727 in gazebo::physics::World::ProcessFactoryMsgs() ()
    at /usr/lib/x86_64-linux-gnu/libgazebo_physics.so.9
#15 0x00007ffff5b1945d in gazebo::physics::World::ProcessMessages() ()
    at /usr/lib/x86_64-linux-gnu/libgazebo_physics.so.9
#16 0x00007ffff5b1ad84 in gazebo::physics::World::Step() ()
    at /usr/lib/x86_64-linux-gnu/libgazebo_physics.so.9
#17 0x00007ffff5b1b1cd in gazebo::physics::World::RunLoop() ()
    at /usr/lib/x86_64-linux-gnu/libgazebo_physics.so.9
#18 0x00007ffff69a166f in  () at /usr/lib/x86_64-linux-gnu/libstdc++.so.6
#19 0x00007ffff5dfd6db in start_thread (arg=0x7fff1ffff700)
    at pthread_create.c:463
#20 0x00007ffff63fc88f in clone ()
    at ../sysdeps/unix/sysv/linux/x86_64/clone.S:95

‌

I think this is because of versions of assimp.
This commit of assimp assimp/assimp@36e53b7#diff-758737a587ca204dec82e47c452466d2
changed the output collada mesh
ftom

<mesh>
  ...
  <vertices id="meshId0-vertices">
    <input semantic="POSITION" source="#meshId0-positions" />
    <input semantic="NORMAL" source="#meshId0-normals" />
  </vertices>
  <polylist count="11" material="defaultMaterial">
    <input offset="0" semantic="VERTEX" source="#meshId0-vertices" />
    <vcount>3 3 3 3 3 3 3 3 3 3 3 </vcount>
    <p>0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 17 18 15 19 20 21 21 22 19 23 24 25 26 27 28 </p>
  </polylist>
</mesh>

to

<mesh>
  ...
  <vertices id="meshId0-vertices">                                                                                
    <input semantic="POSITION" source="#meshId0-positions" />                                                     
  </vertices>                                                                                                     
  <polylist count="11" material="defaultMaterial">                                                                
    <input offset="0" semantic="VERTEX" source="#meshId0-vertices" />                                             
    <input offset="0" semantic="NORMAL" source="#meshId0-normals" />                                              
    <vcount>3 3 3 3 3 3 3 3 3 3 3 </vcount>                                                                       
    <p>0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 17 18 15 19 20 21 21 22 19 23 24 25 26 27 28 </p>              
  </polylist>
</mesh>

This modification may cause index out of range at this line of gazebo.
https://github.com/osrf/gazebo/blob/f57e4e95988c14f99ee48f736253b76ef5309f10/gazebo/common/ColladaLoader.cc#L1548