Limit the size for typed arrays to MaxSmi.

Limit the size for typed arrays to MaxSmi.

R=jkummerow@chromium.org
LOG=Y
BUG=319722
Committed: https://code.google.com/p/v8/source/detail?r=17800

[Dmitry Lomov (no reviews), 2013-11-15 12:08:28]

PTAL.

In tests, we need to allocate a large ArrayBuffer for each test, so we make sure
that we only allocate one per test and disable --stress-opt mode.

[Dmitry Lomov (no reviews), 2013-11-15 12:49:01]

Tests made lighter + fail earlier in TypedArray(length) constructor.

[Jakob Kummerow, 2013-11-15 15:13:08]

LGTM with comments.

https://codereview.chromium.org/73943004/diff/110001/src/runtime.cc
File src/runtime.cc (right):

https://codereview.chromium.org/73943004/diff/110001/src/runtime.cc#newcode944
src/runtime.cc:944: // Returns true if backing stor e was initialized or false
otherwise.
nit: accidental edit

https://codereview.chromium.org/73943004/diff/110001/src/typedarray.js
File src/typedarray.js (right):

https://codereview.chromium.org/73943004/diff/110001/src/typedarray.js#newcode91
src/typedarray.js:91: throw MakeRangeError(length,
"invalid_typed_array_length");
nit: {} please

https://codereview.chromium.org/73943004/diff/110001/test/mjsunit/regress/reg...
File test/mjsunit/regress/regress-319722-ArrayBuffer.js (right):

https://codereview.chromium.org/73943004/diff/110001/test/mjsunit/regress/reg...
test/mjsunit/regress/regress-319722-ArrayBuffer.js:33: for (k = 8; k >= 1 && ab
== null; k = k/2) {
the update step should be "k = k >> 1", otherwise "k >= 1" will never become
false and a low memory condition will lead to an endless loop.

https://codereview.chromium.org/73943004/diff/110001/test/mjsunit/regress/reg...
File test/mjsunit/regress/regress-319722-Float32Array.js (right):

https://codereview.chromium.org/73943004/diff/110001/test/mjsunit/regress/reg...
test/mjsunit/regress/regress-319722-Float32Array.js:30: assertThrows(function ()
{ new Float32Array(maxSize); }, RangeError);
We could combine all of these tests into a single file too, right?

[Dmitry Lomov (no reviews), 2013-11-15 16:03:07]

Thanks for review, all comments addressed, landing.

https://codereview.chromium.org/73943004/diff/110001/src/runtime.cc
File src/runtime.cc (right):

https://codereview.chromium.org/73943004/diff/110001/src/runtime.cc#newcode944
src/runtime.cc:944: // Returns true if backing stor e was initialized or false
otherwise.
On 2013/11/15 15:13:08, Jakob wrote:
> nit: accidental edit

Done.

https://codereview.chromium.org/73943004/diff/110001/src/typedarray.js
File src/typedarray.js (right):

https://codereview.chromium.org/73943004/diff/110001/src/typedarray.js#newcode91
src/typedarray.js:91: throw MakeRangeError(length,
"invalid_typed_array_length");
On 2013/11/15 15:13:08, Jakob wrote:
> nit: {} please

Done.

https://codereview.chromium.org/73943004/diff/110001/test/mjsunit/regress/reg...
File test/mjsunit/regress/regress-319722-ArrayBuffer.js (right):

https://codereview.chromium.org/73943004/diff/110001/test/mjsunit/regress/reg...
test/mjsunit/regress/regress-319722-ArrayBuffer.js:33: for (k = 8; k >= 1 && ab
== null; k = k/2) {
On 2013/11/15 15:13:08, Jakob wrote:
> the update step should be "k = k >> 1", otherwise "k >= 1" will never become
> false and a low memory condition will lead to an endless loop.

No, I think it is fine as it is, because 1/2 = 0.5 < 1

https://codereview.chromium.org/73943004/diff/110001/test/mjsunit/regress/reg...
File test/mjsunit/regress/regress-319722-Float32Array.js (right):

https://codereview.chromium.org/73943004/diff/110001/test/mjsunit/regress/reg...
test/mjsunit/regress/regress-319722-Float32Array.js:30: assertThrows(function ()
{ new Float32Array(maxSize); }, RangeError);
On 2013/11/15 15:13:08, Jakob wrote:
> We could combine all of these tests into a single file too, right?

Done.

[Dmitry Lomov (no reviews), 2013-11-15 16:10:09]

Committed patchset #4 manually as r17798 (presubmit successful).

[Dmitry Lomov (no reviews), 2013-11-15 16:35:17]

On 2013/11/15 16:10:09, Dmitry Lomov (chromium) wrote:
> Committed patchset #4 manually as r17798 (presubmit successful).

Reverted - apparently buffers we try to allocate are too large. Uploaded a new
patch that reduce buffer size and also fix the message id.

[Jakob Kummerow, 2013-11-15 16:36:19]

Patch set 5 LGTM.

[Dmitry Lomov (no reviews), 2013-11-15 16:37:27]

Committed patchset #5 manually as r17800 (presubmit successful).