bc: runtime error eats too much input

[mknos]

• In interactive bc, calling an undefined function does not cause a fatal error but an error is printed
• YYERROR() is called in this situation, but it seems to break something
• It appears that input tokens past the end of current statement are eaten by YYERROR
• I observed this:

ME> x = hey(1) + 2
BC> No function hey() has been defined
ME> x
ME> x
BC> 0
ME> quit

• After the error bc ignores the first time I type x (but when I ask again it politely answers)
• Replacing YYERROR with "last INSTR" causes bc to jump past the final instruction in the current statement
• Instructions in the subsequent statement are not destroyed
• With this patch, the statement after the error is executed as I expect (this matches GNU bc)

ME> x = 5
BC> 5
ME> y = heh(6) + 7
BC> No function heh() has been defined
ME> x
BC> 5
ME> y
BC> 0
ME> quit

[briandfoy]

I see this, but I want to look at it more closely. The innards of bc are quite complicated.

[mknos]

• exec_stmt() can be called recursively, e.g. for handling an if-statement
• In the example I gave, exec_stmt() is not called recursively
• New example below
• The underlying language is a stack, and assigning into x happens last
• Push 10 and 5 onto stack, y() is expected to take its single argument (5) off the stack
• Return value of y() replaces 5 in stack, then multiplies with 10
• Push 2 on stack; then 2 is added to result of multiplication
• In my understanding, calling "last INSTR" discards everything in the current statement (the list called stmt_s)
• So in this example nothing is assigned into x, because the result can't be determined without y() being defined

x = 10 * y(5) + 2
executing statement: $VAR1 = [
          [
            'N',
            10
          ],
          [
            'N',
            5
          ],
          [
            'FUNCTION-CALL',
            'y'
          ],
          [
            '*_'
          ],
          [
            'N',
            2
          ],
          [
            '+_'
          ],
          [
            'V',
            'x'
          ],
          [
            '=V'
          ]
        ];
No function y() has been defined

[briandfoy]

I have not forgotten about this. I should have time in a week to look at this. I don't think there's a problem but I want to understand the situation first.

[briandfoy]

approved