rostest: use AnyMsg in publishtest

[beetleskin]

Right now publishtest only works correctly when a publisher of the given topic exists: If no publisher exists, the retrieval of the topic class blocks until the test times out. This leads to aborted tests when one would expect a failed one with correct error message, stating that no message on the given topic was published. Using rospy.AnyMsg fixes that.

This also makes it impossible to test that a certain topic is not being published, when no publisher exists.

[beetleskin]

Uff .. the failing test is a hztest, without retry argument. This is unfortunate, this never works reliably on CI with shared resources. I could add a retry argument, but this change is not about fixing those tests for CI.

Other instances of this from the recent PRs:

• Add set_rate service #1647
• Another approach: use python-cryptography instead #1606
• .. and then the Jenkins history is cut off (I guess, you get 404 for older Details/test results)

How can I just trigger the build again?

[beetleskin]

How can I just trigger the build again?

https://stackoverflow.com/a/28703197/1880234
.. but now we have this empty commit in there :/

[dirk-thomas]

I am not sure why you closed the ticket. The empty commit is not a problem - at the end the PR will be squash merged anyway. Do you want to reopen this?

Nvm, just saw that you also deleted the fork...

[beetleskin]

messed up my fork, there you go: #1659