Handling of >> and << with negative shift values is not specified/wrong in dart2js?

[DartBot]

This issue was originally filed by jimhug@google.com

---

These two cases behave differently between frog/dartc and the vm in production mode. I'd like to confirm that this is a bug an frog and also decide if this needs to be in the Editor milestone or not.
Kasper - This is assigned to you to confirm the correct behavior and to assign it to that milestone if you choose. I am separating this out from Issue #596 to track the important differences.

void main() {
  var x = 77;
  print(x >> (-1)); // frog,dartc: 0, VM: exception
  print(foo() << (-1)); // frog,dartc: -2147483648, VM: exception
}

The part of this bug that is critical is resolving the expected behavior and whether or not this needs to be moved into the Editor milestone.

[dgrove]

The fact that these aren't well-defined is a dup of issue #1136.

I believe that the milestone issue here is that frog should be calling a function rather than using a straight <</>> unless it can prove both of the following:
  1. lhs of the operation is an integer
  2. rhs of the operation is an integer that is >= 0;

---

Added label.

[DartBot]

This comment was originally written by jimhug@google.com

---

I'm sorry, but the statement that "frog should be calling a function unless..." is not a useful way to specify expected behavior. For this to be a frog bug, it needs to explain what the expect output of the program is. When I went to fix this bug, I compared results to dartc in order to understand what our current best specification of correct behavior of a system with the JS numeric exemption was.

Your rule is one interesting possibility for the correct behavior, but it seems like an awful lot of work for very little value to me. There is a broad range of values for which frog will currently generate different results on the << and >> operators than the VM will.

For x << y, these include y < 0 as you point out. However, they also include y > 30 (unless x == 0) as well as x > 2^30ish for any values of y - including 0. For each value of y between 0 and 30, there is a different range of values of x that will and will not match the VM.

In order to fix this bug in frog I need someone to clearly specify the correct behavior - as I'm hoping to get out of issue #1136. Until that is resolved, I would be prefer to just mark this as blocked on that issue. However, I'm waiting to see if Kasper feels differently and would like to give me a precise specification for how these operators should work on a Dart implementation that compiles to JS if he feels it needs to be fixed as part of the editor milestone.

[kasperl]

Jim has already started working on improving the testing situation in this CL: http://codereview.chromium.org/9149005/.

[kasperl]

Moving this out of Area-Frog and putting it into Area-Language for the lack of a better place. One thing we could say is that we'll use JavaScript semantics for shifts when both the LHS and the RHS are numbers.

---

Removed Area-Frog label.
Added Area-Language label.

[anders-sandholm]

Added this to the M1 milestone.

[gbracha]

Moving out of language per discussion with Kasper.

---

Removed Area-Language label.
Added Area-Dart2JS label.

[kasperl]

We've decided to stick with unsigned 32-bit int output from bitwise operation and no range checks for the int inputs for M1.

---

Removed this from the M1 milestone.
Added this to the Later milestone.
Removed Priority-Critical label.
Added Priority-Medium label.
Changed the title to: "Handling of >> and << with negative shift values is not specified/wrong in dart2js?".

[kasperl]

Removed this from the Later milestone.

[kasperl]

Added this to the Later milestone.

[lrhn]

Issue #9400 has been merged into this issue.

[kasperl]

Added Triaged label.

[kasperl]

Added TriageForM5 label.

[kasperl]

Removed TriageForM5 label.

[kasperl]

Issue #11521 has been merged into this issue.

[lrhn]

Shift is now specified (or will soon be) and implemented to not allow negative shifts.

---

Added Fixed label.

[DartBot]

This comment was originally written by jdav...@pcprogramming.com

---

what version does this roll out in?

[floitschG]

The negative shift check has been implemented a long time ago.