Bug 46990 – TypedArray - NaNs converted to 0 in Float32 array

• Status: RESOLVED
• Resolution: FIXED
• Component: WebGL

[Joshua Bell]

var n = 1 / "abc";
>> NaN

new Float32Array([n])[0];
>> 0, expecting NaN

Other edge cases (positive/negative infinities, negative zero) behave correctly:

new Float32Array([1/0])[0];
>> Infinity

new Float32Array([-1/0])[0];
>> -Infinity

var negzero = -0;
function isnegzero(x) { return 1/x == -Infinity; }
isnegzero(new Float32Array([negzero])[0]);
>> true

[Adrienne Walker]

Created attachment 77587 [details]
Patch

[Adrienne Walker]

(In reply to comment #1)
> Created an attachment (id=77587) [details]
> Patch

I will update the Khronos conformance test once this lands.

[WebKit Commit Bot]

Comment on attachment 77587 [details]
Patch

Rejecting attachment 77587 [details] from commit-queue.

Failed to run "['./Tools/Scripts/webkit-patch', '--status-host=queues.webkit.org', '--bot-id=cr-jail-4', 'build-and-test', '--no-clean', '--no-update', '--test', '--non-interactive']" exit_code: 2
Last 500 characters of output:
................................................
fast/box-shadow ......
fast/box-sizing ....
fast/canvas .....................................................................................................................................
fast/canvas/webgl .......
fast/canvas/webgl/array-unit-tests.html -> failed

Exiting early after 1 failures. 6210 tests run.
126.95s total testing time

6209 test cases (99%) succeeded
1 test case (<1%) had incorrect layout
4 test cases (<1%) had stderr output

Full output: http://queues.webkit.org/results/7198304

[WebKit Commit Bot]

Comment on attachment 77587 [details]
Patch

Rejecting attachment 77587 [details] from commit-queue.

Failed to run "['./Tools/Scripts/webkit-patch', '--status-host=queues.webkit.org', '--bot-id=cr-jail-3', 'build-and-test', '--no-clean', '--no-update', '--test', '--non-interactive']" exit_code: 2
Last 500 characters of output:
................................................
fast/box-shadow ......
fast/box-sizing ....
fast/canvas .....................................................................................................................................
fast/canvas/webgl .......
fast/canvas/webgl/array-unit-tests.html -> failed

Exiting early after 1 failures. 6210 tests run.
127.07s total testing time

6209 test cases (99%) succeeded
1 test case (<1%) had incorrect layout
4 test cases (<1%) had stderr output

Full output: http://queues.webkit.org/results/7324244

[Adrienne Walker]

Created attachment 77946 [details]
Patch

[Adrienne Walker]

Fixed test error on OSX because Int32Array([NaN]) was converting the value improperly.  Also, refactored the test so that optimized code paths would get tested too.

[Kenneth Russell]

Comment on attachment 77946 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=77946&action=review

Looks good overall; one question. If there's an error in the test, just feel free to fix it upon landing the patch.

> LayoutTests/fast/canvas/webgl/array-unit-tests.html:618
> +      for (var i = 1; i < array.length; ++i)

Did you mean to start this iteration at 1?

> LayoutTests/fast/canvas/webgl/array-unit-tests.html:622
> +      for (var i = 1; i < array.length; ++i)

Same question here.

[Adrienne Walker]

Comment on attachment 77946 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=77946&action=review

>> LayoutTests/fast/canvas/webgl/array-unit-tests.html:618
>> +      for (var i = 1; i < array.length; ++i)
> 
> Did you mean to start this iteration at 1?

Thanks for catching that.  I must have missed those two loops when refactoring some code.

[Adrienne Walker]

Committed r75167: <http://trac.webkit.org/changeset/75167>

[WebKit Review Bot]

http://trac.webkit.org/changeset/75167 might have broken Qt Linux Release