Fix array-out-of-bounds in TypedDataView::ElementSizeInBytes.

Fix array-out-of-bounds in TypedDataView::ElementSizeInBytes.

Also add index bounds checking for the lookup table.

R=iposva@google.com

Committed: https://code.google.com/p/dart/source/detail?r=44101

[koda, 2015-02-24 21:47:38]

Note that DEBUG mode also didn't catch this one. I'm working on a suitable unit
test.

FYI, it was called from here; I'm still trying to estimate the likelihood that
we were actually bitten by it:

https://code.google.com/p/dart/source/browse/branches/bleeding_edge/dart/runt...

[koda, 2015-02-24 23:06:22]

Not adding a unit test in this CL. The code will be exercised by
dart_api_impl_test after this CL: https://codereview.chromium.org/957483002/

PTAL

[koda, 2015-02-25 23:48:46]

On 2015/02/24 23:06:22, koda wrote:
> Not adding a unit test in this CL. The code will be exercised by
> dart_api_impl_test after this CL: https://codereview.chromium.org/957483002/
> 
> PTAL

Ping. This is blocking https://codereview.chromium.org/957483002/

[Ivan Posva, 2015-02-25 23:52:44]

LGTM with comment.

-Ivan

https://codereview.chromium.org/949343002/diff/80001/runtime/vm/object.h
File runtime/vm/object.h (right):

https://codereview.chromium.org/949343002/diff/80001/runtime/vm/object.h#newc...
runtime/vm/object.h:7039: static const intptr_t kNumElementSizes = 14;
Where does 14 come from?

[koda, 2015-02-25 23:57:59]

https://codereview.chromium.org/949343002/diff/80001/runtime/vm/object.h
File runtime/vm/object.h (right):

https://codereview.chromium.org/949343002/diff/80001/runtime/vm/object.h#newc...
runtime/vm/object.h:7039: static const intptr_t kNumElementSizes = 14;
On 2015/02/25 23:52:44, Ivan Posva wrote:
> Where does 14 come from?

That's the number of elements in the array; see object.cc. Cannot use ARRAY_SIZE
here since the definition is not available, and we want this to be inlined. Do
you prefer I make a (redundant) enum instead?

[Ivan Posva, 2015-02-26 00:15:44]

https://codereview.chromium.org/949343002/diff/80001/runtime/vm/object.h
File runtime/vm/object.h (right):

https://codereview.chromium.org/949343002/diff/80001/runtime/vm/object.h#newc...
runtime/vm/object.h:7039: static const intptr_t kNumElementSizes = 14;
On 2015/02/25 23:57:59, koda wrote:
> On 2015/02/25 23:52:44, Ivan Posva wrote:
> > Where does 14 come from?
> 
> That's the number of elements in the array; see object.cc. Cannot use
ARRAY_SIZE
> here since the definition is not available, and we want this to be inlined. Do
> you prefer I make a (redundant) enum instead?

Isn't it "(kTypedDataFloat64x2ArrayCid - kTypedDataInt8ArrayCid) + 1" or
something similar?

[koda, 2015-02-27 18:21:57]

https://codereview.chromium.org/949343002/diff/80001/runtime/vm/object.h
File runtime/vm/object.h (right):

https://codereview.chromium.org/949343002/diff/80001/runtime/vm/object.h#newc...
runtime/vm/object.h:7039: static const intptr_t kNumElementSizes = 14;
On 2015/02/26 00:15:44, Ivan Posva wrote:
> On 2015/02/25 23:57:59, koda wrote:
> > On 2015/02/25 23:52:44, Ivan Posva wrote:
> > > Where does 14 come from?
> > 
> > That's the number of elements in the array; see object.cc. Cannot use
> ARRAY_SIZE
> > here since the definition is not available, and we want this to be inlined.
Do
> > you prefer I make a (redundant) enum instead?
> 
> Isn't it "(kTypedDataFloat64x2ArrayCid - kTypedDataInt8ArrayCid) + 1" or
> something similar?

Done.

Sure. Until we add another type :) In any case of size mismatch, either the
compiler will complain about too few entries, or the assert will fire on a zero
entry.

[koda, 2015-02-27 18:52:25]

Committed patchset #6 (id:100001) manually as r44101 (presubmit successful).