Unneeded printStackTrace

At https://github.com/zeromq/jeromq/blob/713a3e0f0e940497cc58b0d6dcc0f6873e35301e/src/main/java/zmq/Signaler.java#L137

We have ZMQ that print a stack trace to stderr. I don’t think a library should be allowed to directly do that. Especialy when the exception is as quite normal java.nio.channels.ClosedByInterruptException and is thrown during . a plain: zpoller.poll(-1L);

[fredoboulo]

I would be more radical:

• If the exception happens and is handled, there is no need for a stacktrace.
• It the exception happens and is unhandled, that means some devs need to be performed to handle the situation.
• If the exception does not happen, an assert (false) is even better as it enforces the behaviour of the dead branch of the code.

I did not see occurrences of this stacktrace (and the others of this file, but was not keen to enforce it drastically so I left these traces (several of them are in this file as well).

I agree with you: No stacktrace. If the library handles the situation.

[daveyarwood]

I still see 56 instances of printStackTrace in the repo, 18 of which are in the library code (as opposed to in tests, where it might be more feasible to do it).

I agree with @fredoboulo about taking a more extreme approach to this. I don’t think there’s any reason why we should print stacktraces to STDERR and then continue running the program. We should handle all exceptions or let them explode.