Serious bug identified in SmartDashboard/NetworkTables -- robot hangs

[virtuald]

Originally Posted by NotInControl

I was just writing to make sure I understand the bug because we use Java and it was unclear if this problem was just for C++, on the client side/ ther server side, both…

The problem will affect both the client and server side.

I believe the problem is most severe on C++, but that similar problems may affect the java side, as others have reported less severe problems in Java. I've only glanced at the java version recently, and a cursory look shows the java code does not hold the lock the same way that the C++ code, but my expectation is that there is a similar bug somewhere, given that it's written by the same author.

Based on what I read it appears the crux of the problem is that although NT is multithreaded, it holds on to a lock during a write sequence. A write sequence which also blocks and keeps the lock if the write fails. The robot thread uses this same lock to push data, thus causing the hang on the robot. Is this all correct?

Yup.

My question is the code to obtain the lock on the robotside in NT or in SmartDash? If you were to call the smartDashboard putXXX methods in a separate thread, would that not at least prevent the hang?

That should prevent the hang.