osrf-migration
commented
11 years ago Original comment by Steve Peters (Bitbucket: Steven Peters, GitHub: scpeters).
A workaround is to change the order of tests in physics.cc
diff -r 519a3b14f23af79096868d0e037046137f1572ac test/regression/physics.cc
--- a/test/regression/physics.cc Mon Apr 22 16:07:44 2013 -0700
+++ b/test/regression/physics.cc Tue Apr 23 09:40:48 2013 -0700
@@ -33,6 +33,13 @@
public: void CollisionFiltering(const std::string &_physicsEngine);
};
+// As a workaround for issue #660, this test is placed first.
+// When issue #660 is resolved, this can be moved back.
+TEST_F(PhysicsTest, RevoluteJointODE)
+{
+ RevoluteJoint("ode");
+}
+
////////////////////////////////////////////////////////////////////////
// EmptyWorld:
// Load a world, take a few steps, and verify that time is increasing.
@@ -921,11 +928,6 @@
}
}
-TEST_F(PhysicsTest, RevoluteJointODE)
-{
- RevoluteJoint("ode");
-}
-
#ifdef HAVE_BULLET
TEST_F(PhysicsTest, RevoluteJointBullet)
{
Original report (archived issue) by Steve Peters (Bitbucket: Steven Peters, GitHub: scpeters).
The Physics.RevoluteJoint tests are failing since they are unable to download the double_pendulum_with_base model. Some debugging indicated that the ModelDatabase thread gets killed after the first test (EmptyWorldODE in this case) and the ModelDatabase::modelCache variable is cleared, so that no new models can be downloaded. I don't notice this in testing on my local machine because I have a local copy of gazebo_models in my path.
Currently
common::ModelDatabase::Instance()->Fini();destroys the ModelDatabase thread and is called in World::Stop. Since the ModelDatabase thread is created in its constructor, it never gets recreated. Perhaps creating the thread in ModelDatabase::Init?