Sometimes, when something won't even compile, I have try to compile it a second time - and get the exact same error message(s) - before my brain will accept there is a problem and perhaps begin to see the cause.

Somewhere in my head must be a gremlin who says, "No! It is the compiler that is wrong!"

The same can apply to semantic errors (the code is valid but does not do what it was intended to do) but those take longer to track down. To make the trick work, the debugging output has to be in just the right place in order to print proof of the wrong logic and then do the same on the next run, preferably in under a minute, so that I can begin to see the error.