Fix bond tests

[mjcarroll]

Fix unit tests for CI.

Signed-off-by: Michael Carroll [email protected]

[hidmic]

@mjcarroll why the change? why the difference with bondpy?

[mjcarroll]

The WaitingForSister state doesn't have an implementation for HeartbeatTimeout as you can see here:

bond_core/bondcpp/include/bondcpp/BondSM_sm.h

Lines 67 to 79 in 5a79fec

	class SM_WaitingForSister :
	public SM_Default
	{
	public:
	SM_WaitingForSister(const char *name, int stateId)
	: SM_Default(name, stateId)
	{};
	void ConnectTimeout(BondSMContext& context);
	void Die(BondSMContext& context);
	void SisterAlive(BondSMContext& context);
	void SisterDead(BondSMContext& context);
	};

This was causing a segfault in the remoteNeverConnects test:

bond_core/test_bond/test/test_callbacks_cpp.cpp

Lines 98 to 107 in fb0fa11

	TEST_F(TestCallbacksCpp, remoteNeverConnects)
	{
	auto nh2 = rclcpp::Node::make_shared("test_callbacks_cpp_2");
	std::string id2 = genId();
	bond::Bond a1(TOPIC, id2, nh2);
	a1.start();
	EXPECT_FALSE(a1.waitUntilFormed(rclcpp::Duration(5.0s)));
	EXPECT_TRUE(a1.waitUntilBroken(rclcpp::Duration(10.0s)));
	}

I don't believe that this impacts behavior, just guards against an undefined condition that wasn't previously exposed from the unit tests.

[hidmic]

I'm not super familiar with bond's architecture, but if a heartbeat timeout can occur while waiting for sister, should said state handle that condition (even if ignoring it)? By implementing HeartbeatTimeout I mean. That applies to both bondcpp and bondpy (where I don't an if clause like this one).

[wjwwood]

@mjcarroll bump (this is in the list of things to triage for Galactic)

[sloretz]

IIUC ros2 bondcpp has this new method start() that resets the heartbeat timer, while kinetic-devel bondcpp only reset the heartbeat timer the first time someone called Heartbeat. It probably wasn't possible for the WaitingForSister state to see HeartbeatTimeout in ROS 1.

Maybe the heartbeat timer should be reset only when the connection is formed? I also agree with @hidmic that WaitingForSister should implement that callback too, but the implementation should be a warning or error that the callback is unexpected.

[sloretz]

@mjcarroll is this still targeting initial Galactic release, or is patch 1 more likely?

[mjcarroll]

Probably patch 1 at this point.

[mjcarroll]

Going to backport some of the fixes from #87 in place of this.