Add delay when retrying tests involving the CLI daemon#459
Add delay when retrying tests involving the CLI daemon#459jacobperron merged 1 commit intomasterfrom
Conversation
wjwwood
left a comment
There was a problem hiding this comment.
lgtm
A little theory behind this change, even if you ensure the fixure is setup (similar to what was proposed in #454) you are still racing the discovery. Even if, in the case of ros2 action info tests, you created a client and waited for the server to be "available" it may not appear available to the daemon until after that. Also, you have to wait for the client you used to check this to disappear (another discovery race).
In this solution it will finish immediately if the discovery is "immediate", but will be robust to some discovery delays with the daemon (on the order of 5 seconds in most cases). Before this change, the reties happened quickly, so not much time was given to wait for discovery delays to be resolved.
This is to give time for discovery to happen between the daemon node and the test fixture nodes. Signed-off-by: Jacob Perron <jacob@openrobotics.org>
fd405c3 to
da4a40a
Compare
|
rebased on master. CI is here: ros2/launch#390 (comment) |
This is to give time for discovery to happen between the daemon node and the test fixture nodes.
Alternative to #454
Depends on ros2/launch#390