frog handling null incorrectly

[dgrove]

svn r1844

There are a variety of cases where frog is handling null incorrectly. A few examples are below:

f() => null;

main() {
print (~f());
}

---

f() => null;

main() {
print(f());
}

---

running the VM produces an error in both cases:

NullPointerException
 0. Function: 'Object.noSuchMethod' url: 'bootstrap' line:315 col:3
 1. Function: '::.main' url: '/Users/dgrove/repo/dart-bleeding/dart/frog/y.dart' line:4 col:7

running frog produces "0" in both cases.

Another example:

f() => null;

main() {
print(f() + 7);
}

frog produces "7", the VM throws an exception.

[rakudrama]

When I run the second one on the VM, it prints 'null', which is the only sane behaviour.

For
    f() => null;
    main() => print(~f());
I see the VM throws NPE, Frog prints '-1'

I am running with --enable_type_checks so I expect correct behaviour. (Without --enable_type_checks you might argue that the NPE is a kind of type check on the receiver, so anything goes.)

This example has no exceptions to the issue:
    foo([x]) => '$x';
    main() => Expect.equals('null', foo());
Expect.equals(expected: <null>, actual: <undefined>) fails.
...
I see 'undefined' in interpolated strings fairly often.

[dgrove]

Added Triaged label.

[dgrove]

Added this to the FrogEditor milestone.

[jmesserly]

Yeah, I would expect the negative tests to fail with --enable_type_checks. I'm kind of surprised they aren't working already. I think it happens because operators aren't correctly dispatched as "var".

Regarding the toString issue--I'm not sure what to do there. We either need to give up on undefined being a synonym for null, or generate uglier string interp/concat and toString code.

[jmesserly]

btw, the toString issue should be addressed in this separate bug: http://code.google.com/p/dart/issues/detail?id=618

[jmesserly]

BTW, I think the fix for:
  http://code.google.com/p/dart/issues/detail?id=596

...will fix this too.

[jmesserly]

Marked this as being blocked by #596.

[DartBot]

This comment was originally written by jimhug@google.com

---

Also for the record, I tried these tests in dartc. Dartc does correctly handle the "f() + 7" case; however, it is equally broken as frog on the "7 + f()" case - although dartc seems to return NaN instead of 7. Nevertheless, we will fix this as part of the editor milestone.

---

cc @kasperl.

[DartBot]

This comment was originally written by jimhug@google.com

---

Set owner to jimhug@google.com.
Added Started label.

[DartBot]

This comment was originally written by jimhug@google.com

---

r3505 fixes the specific issues reported in this bug and generally improves null handling in frog.

---

Added Fixed label.