Improved joint warnings for Single and MultiDofJoints

[mxgrey]

Similar to the previous changes for MultiDofJoints, I've standardized and improved the warnings for SingleDofJoints.

This also fixes a bug that caused warning messages when a zero-valued command was set for passive actuators.

[jslee02]

We could consider to use std::endl instead of \n for new line. The difference is that std::endl causes flushing of the output buffer while \n doesn't.

In general, \n would be good for performance by avoiding to flush everytime. However, when we want to know where exactly the message is printed, std::endl would be better. For example, unittest prints warning messages after all the unittests are done so that we don't know which unittest caused which warning messages.

[mxgrey]

In general, \n would be good for performance by avoiding to flush everytime.

This is why I prefer \n. I'd rather trust the computer to know when a good time to flush is, or trust the user to flush when it's good for them, instead of assuming that messages should be displayed immediately.

By default, dtwarn and dterr are streaming to std::cerr, and the default behavior of std::cerr is to not do any buffering (i.e. flush immediately), so the only way dtwarn and dterr would not flush right away is if the user intentionally altered that behavior, in which case I don't think we should be forcing it to flush, because the user is probably doing their own flushing management.

[mxgrey]

For example, unittest prints warning messages after all the unittests are done so that we don't know which unittest caused which warning messages.

I'm guessing this is the fault of gtest. One thing we could consider is explicitly flushing std::cerr at the end of each test, if gtest is buffering cerr for some reason.

[jslee02]

I see. Maybe it was happened for dtmsg which is streaming to std::cout. I'll double check.

[mxgrey]

It's important to note that cerr and cout are completely asynchronous, so messages that are sent out at the same time to each could end up on the terminal in wildly different places. In theory, cerr messages should make it to the terminal before cout messages, but in practice I don't think that's always been the case for me. I'm pretty sure some of the behavior depends on the terminal program itself.

[jslee02]

The print order of gtest and dtmsg/dtwarn/dterr seems random but mostly in the right order. So let's leave flushing management to the user as you said.