Add kHeaderSize constant to SeqString.

Add kHeaderSize constant to SeqString.

This prevents potential misuse of SeqString::kHeaderSize as in the
case of live byte counting in incremental marking stub. All stubs
picked up the undefined size constant SeqString::kHeaderSize, thus
the computed size of all strings was off by two pointers slots.

R=lrn@chromium.org
BUG=v8:1672
TEST=mjsunit/object-seal.js,...

Committed: http://code.google.com/p/v8/source/detail?r=9349

[Michael Starzinger, 2011-09-20 07:42:57]

PTAL.

[Lasse Reichstein, 2011-09-20 07:54:30]

I don't see any change to the incremental marking stub?
(I'm also working on fixes to the live-byte count. I'll upload a CL "soon", so
let's compare).

[Lasse Reichstein, 2011-09-20 07:57:11]

http://codereview.chromium.org/7971009/diff/1/src/objects.h
File src/objects.h (right):

http://codereview.chromium.org/7971009/diff/1/src/objects.h#newcode6132
src/objects.h:6132: static const int kAlignedSize =
POINTER_SIZE_ALIGN(kHeaderSize);
I can see pros and cons with moving this here.
It locks the layouts of ASCII/TwoByte strings to be the same ... but it
simplifies things in the places where we assume that they are the same. I think
the latter is quite prevalent, so the change is good.
Try to check for ASSERT_EQ(SeqAsciiString::kHeaderSize,
SeqTwoByteString::kHeaderSize) or similar comparions that are no longer needed
(they should probably compare alinged size, but I bet most don't).

[Michael Starzinger, 2011-09-20 08:19:08]

Added new patch set.

http://codereview.chromium.org/7971009/diff/1/src/objects.h
File src/objects.h (right):

http://codereview.chromium.org/7971009/diff/1/src/objects.h#newcode6132
src/objects.h:6132: static const int kAlignedSize =
POINTER_SIZE_ALIGN(kHeaderSize);
On 2011/09/20 07:57:11, Lasse Reichstein wrote:
> I can see pros and cons with moving this here.
> It locks the layouts of ASCII/TwoByte strings to be the same ... but it
> simplifies things in the places where we assume that they are the same. I
think
> the latter is quite prevalent, so the change is good.
> Try to check for ASSERT_EQ(SeqAsciiString::kHeaderSize,
> SeqTwoByteString::kHeaderSize) or similar comparions that are no longer needed
> (they should probably compare alinged size, but I bet most don't).

Done.

[Lasse Reichstein, 2011-09-20 08:42:01]

LGTM, but the description still talks about a change to the incremental marking
stub, which doesn't seem to be here.

http://codereview.chromium.org/7971009/diff/3004/src/objects.h
File src/objects.h (right):

http://codereview.chromium.org/7971009/diff/3004/src/objects.h#newcode6132
src/objects.h:6132: static const int kAlignedSize =
POINTER_SIZE_ALIGN(kHeaderSize);
Do we ever use kAlignedSize any more?
Since we made the string length a Smi, it's always the same as kHeaderSize (even
on x64, where previously it was kIntSize less), so we could remove it now.

[Michael Starzinger, 2011-09-20 16:07:39]

Added new patch set. Updated the description to better reflect what this CL
changes. Enabled the rest of the test cases associated with the given issue.

http://codereview.chromium.org/7971009/diff/3004/src/objects.h
File src/objects.h (right):

http://codereview.chromium.org/7971009/diff/3004/src/objects.h#newcode6132
src/objects.h:6132: static const int kAlignedSize =
POINTER_SIZE_ALIGN(kHeaderSize);
On 2011/09/20 08:42:01, Lasse Reichstein wrote:
> Do we ever use kAlignedSize any more?
> Since we made the string length a Smi, it's always the same as kHeaderSize
(even
> on x64, where previously it was kIntSize less), so we could remove it now.

Done.