Error Recovery in Bottom-Up (2)

• If the parser encounters an error, it will pop the stack until it gets into a state, where error is legal.
• At this point it shifts error onto the stack.
• Then, the input tokens are skipped, until the next input token is one that can legally follow the new state.
• This scheme is very dependent on a good choice of error productions.
• Assume a production Statement = error ";"
  • The parser encounters error inside a statement. It will pop the stack until it expects a statement.
  • At this point it shifts error onto the stack.
  • Then, the input tokens are skipped, until ";" is found.