Fix new ArrayBuffer(length) for various input

[glistening]

Currently new ArrayBuffer(length) does not conform to ES2015.

For example, new ArrayBuffer(length) should not throw RangeError
for length = NaN, undefined, negative number, floating point, and so on.

JerryScript-DCO-1.0-Signed-off-by: Sanggyu Lee [email protected]

[glistening]

@jiangzidong Could you review this patch?

[jiangzidong]

@glistening I dont think ES2015 allow the length could be NaN, undefined, negative number, floating point.

http://www.ecma-international.org/ecma-262/6.0/#sec-arraybuffer-length see step 5

If SameValueZero(numberLength, byteLength) is false, throw a RangeError exception.

If the length is NaN, undefined, negative number, floating point. they will fail the samevaluezero check

But interesting part is V8 supports NaN, undefined, negative number (-1, 0), floating point. I don't know why.

Correct me if I miss sth in the spec.

[glistening]

@jiangzidong Yes, it is hidden in step 3. Please see ToLength in step 3. Please see ToInteger also, which is called from ToLength. It specifies the handling of NaN, negative number, and so on. Test262 checks this inputs also.

[jiangzidong]

@glistening Yes, I know that ToInteger will turn "negative number, NaN and floating number" into non-negative integer. But then it will fail the step5.
for example, var a = new ArrayBuffer(3.4);

2. Let numberLength be ToNumber(length).

the numberLength is 3.4

3. Let byteLength be ToLength(numberLength).

the byteLength is 3

5. If SameValueZero(numberLength, byteLength) is false, throw a RangeError exception.

3 !== 3.4, so it will throw RangeError .

I don't know why test262/v8/spidermonkey don't follow it.

[glistening]

@jiangzidong Thank you for correcting me. ES2015 says it should throw.

Test262 and major engines (v8, spidermonkey and JavaScriptCore) seems to follow ES2017, not ES2015. ES2017 uses ToIndex to describe the expected behaviour of ArrayBuffer(length). You can see it in test262 test case comment also.

esid: sec-arraybuffer-length
description: >
The length parameter is converted to a value numeric index value.
info: |

ArrayBuffer( length )

1. If NewTarget is undefined, throw a TypeError exception.
2. Let byteLength be ? ToIndex(length).
3. Return ? AllocateArrayBuffer(NewTarget, byteLength).

ToIndex( value )

1. If value is undefined, then
   a. Let index be 0.
2. Else,
   a. Let integerIndex be ? ToInteger(value).
   b. If integerIndex < 0, throw a RangeError exception.
   c. Let index be ! ToLength(integerIndex).
   d. If SameValueZero(integerIndex, index) is false, throw a RangeError exception.
3. Return index.

I will raise this as Issue to get opinion.

[jiangzidong]

@glistening Wow, I didn't notice that it's ES2017. Thanks for pointing out.

[zherczeg]

LGTM

[zherczeg]

(I wouldn't create subsets for such differences, just use the latest standard)

[jiangzidong]

if there is no expection, we dont need to check the name. just

var a = new ArrayBuffer(5.5);
assert(a.byteLength === 5);

[glistening]

@jiangzidong I've simplified 24.01.02-006.js ~24.01.02-012.js.

[jiangzidong]

can you also give a case where the length is float and below -1?

[glistening]

@jiangzidong Now, 24.01.02-012.js covers two more cases that the length is float and below -1.

[jiangzidong]

-1.9 and -2.2 should throw error, see ToIndex 2.b

b. If integerIndex < 0, throw a RangeError exception.

[glistening]

@jiangzidong Oh, You're right. I will fix right now.

[glistening]

@jiangzidong I've updated PR.

[jiangzidong]

LGTM, thanks!