issue-#191 Compiler trap by erroneous program

[Zinn]

The compiler produce a NIL dereference (read) trap instead of an error message

Code: Select all

MODULE TestCompilerError;

   CONST
      length = 5;

   PROCEDURE Name* (): POINTER TO ARRAY OF CHAR;
      VAR string: POINTER TO ARRAY OF CHAR;
   BEGIN
      NEW(string, LEN length + 1);
      RETURN string
   END Name;

END TestCompilerError.


Reported by Rainer Neubauer.

[Josef Templ]

I can confirm the bug and I have added an issue for this.
see https://redmine.blackboxframework.org/issues/191.

Helmut, can you rename this topic to the usual form with the issue-#191 prepended?

- Josef

[luowy]

The easiest way is guard the "node" not NIL, I didn't find out what side effects it has.
DevCPB.StParN

Code: Select all

                 ....
         ELSIF f IN intSet THEN
            IF f = Int64 THEN Convert(x, DevCPT.int32typ) END;
            IF (x.class = Nconst) & (x.conval.intval <= 0) THEN err(63) END;
            node := p.right;
            IF node # NIL THEN   (*+++*)
               WHILE node.link # NIL DO node := node.link END;
               node.link := x; p.typ := p.typ.BaseTyp
            END   (* +++ *)
         ELSE err(111)
         END
                ....


luowy

[Josef Templ]

thanks, luowy, I will look into this tomorrow.

- Josef

[Josef Templ]

I have fixed the trap as proposed by luowy.
In addition, I have added trivial error recovery code:
instead of appending to a non-empty list in the error case, the first element is inserted into the list.
The error message has been generated before.

For the changes see https://redmine.blackboxframework.org/projects/blackbox/repository/diff?utf8=%E2%9C%93&rev=a52a74d8cec9f3fd62950638258aed012de051be&rev_to=a940ecb3063c8387b57c69e4496ea14e641e5c4b


- Josef

[luowy]

the revised code works well, but the title is a bit fuzzy, how about "compiler trap by NEW procedure" ?

[Josef Templ]

the revised code works well, but the title is a bit fuzzy, how about "compiler trap by NEW procedure" ?

I agree that the title could be improved.
However, it should also not remove the fact that the trap is only a follow-up error after another error.
This is hard to express in only a few words.

The precise title would be something like this:
"Compiler trap after invalid first dimension of dynamic array NEW"

- Josef

[Josef Templ]

any suggestions about the title?

Otherwise this topic is also ready for voting.

- Josef