turtlebot/3ea2c30: liveness -> harmless?

[gavanderhoorn]

turtlebot/3ea2c30 is about an env hook (basically a Bash script 'plugin') trying to find a directory which doesn't always exist (as it's a directory from a different package, which isn't listed as a dependency of the package in which the failing env hook is located):

robust/turtlebot/3ea2c30/3ea2c30.bug

Lines 4 to 14 in 7e3932c

	title: Environment Hooks Introduce Undocumented Dependency
	description: >
	ROS packages can make use of environment hooks -- shell scripts
	that are executed, for instance, to set up environment variables.
	In ROS Indigo, the core Turtlebot package introduced an
	environment hook that depended on another package that was not
	listed as a dependency in the package manifest. This caused error
	messages to be printed to the terminal if the user did not have
	the package installed.
	The fix was to move the faulty parts of the environment hook to
	the package they depended on.

This is classified as SOFTWARE:LIVENESS.

However, the net effects of that env hook failing are essentially:

1. an annoying warning on the console every time a user sources their setup.bash (ie: activates a workspace)
2. the TURTLEBOT_MAP_FILE environment variable not getting set

but the base functionality of the package itself (ie: turtlebot_bringup) is not affected. The robot can still be controlled, no software will fail to work, except some other package which might use this variable (some .launch files in turtlebot_navigation).

Do we feel this is sufficient for a LIVENESS label?

---

Edit: and the env hook also only sets a "sane default".

[git-afsantos]

That does not sound like a LIVENESS failure, at all.
If not sure if there is a right label for it, but I would guess that the SOFTWARE is perfectly capable of making progress, despite the fault.

[gavanderhoorn]

Yeah, I'm getting a bit confused about our definition of liveness. The following bugs all have LIVENESS:

• confidential/2688e7a/2688e7a.bug
• confidential/332f09f/332f09f.bug
• confidential/c5dc9de/c5dc9de.bug
• geometry2/12605ab/12605ab.bug
• geometry2/15b2e3c/15b2e3c.bug
• geometry2/245b02b/245b02b.bug
• geometry2/566092b/566092b.bug
• geometry2/7284788/7284788.bug
• geometry2/74f0c66/74f0c66.bug
• kobuki/03660af/03660af.bug
• kobuki/054c753/054c753.bug
• kobuki/38dce2a/38dce2a.bug
• kobuki/3c4c399/3c4c399.bug
• kobuki/6e748c1/6e748c1.bug
• kobuki/acd5bfb/acd5bfb.bug
• mavros/29af3a3/29af3a3.bug
• mavros/86255ba/86255ba.bug
• motoman/377d7be/377d7be.bug
• ros_comm/eab0d3c/eab0d3c.bug
• turtlebot/3ea2c30/3ea2c30.bug
• universal_robot/bd1fce5/bd1fce5.bug

I've spot-checked these, and some don't describe runtime issues.

Could it be we've inflated liveness to include everything which could lead to loss of functionality?

[gavanderhoorn]

Tentative conclusion: see whether SOFTWARE:NETWORK should be reintroduced.

@ChrisTimperley @wasowski

[hsd-dev]

All these bugs listed above seem to have SYSTEM:LIVENESS in the excel sheet. These are not SOFTWARE:LIVENESS