Recovering from Errors in Subrules

[alerosmile]

I’m trying to understand what’s going on during error reporting and recovery. For this I’m debugging the sample from the ANTLR book with the following input (also from the book):

class T {
  int x;
  y;;;
  int z;
}

As expected an described in the book, the parser resynchronizes and stays within the loop because it identified the following variable z. If I put the ‘y’ line to the top of the class, the parser bails out and identifies none of the variables. I was wondering why and found that if I remove the Exception from the DefaultErrorStrategy (and therefore falling through to the next case), the parser resynchronizes and identifies x and z.

This behavior is described in the paragraph ‘Recovering from Errors in Subrules’ (Subrule start / Looping subrule continuation test) but I do not understand why it is like that. Why does the parser not try to resynchronize the same way at the start of the loop?

Thanks a lot.