Fix Windows build by removing an overly strict assertion.

Fix Windows build by removing an overly strict assertion.

The Windows compiler may align members so that the assertion
about the size of RawCode and the offset data_ does not match anymore.

TBR=iposva@google.com

Committed: https://code.google.com/p/dart/source/detail?r=36993

[Florian Schneider, 2014-06-04 13:20:06]

Committed patchset #1 manually as r36993 (presubmit successful).

[Ivan Posva, 2014-06-04 15:53:14]

https://codereview.chromium.org/312153004/diff/1/runtime/vm/object.h
File runtime/vm/object.h (left):

https://codereview.chromium.org/312153004/diff/1/runtime/vm/object.h#oldcode3631
runtime/vm/object.h:3631: ASSERT(sizeof(RawCode) == OFFSET_OF(RawCode, data_));
There are too many things riding on the assertion that sizeof(RawCode) gives us
the correct information. We will have to solve this differently.

[Florian Schneider, 2014-06-06 09:13:03]

https://codereview.chromium.org/312153004/diff/1/runtime/vm/object.h
File runtime/vm/object.h (left):

https://codereview.chromium.org/312153004/diff/1/runtime/vm/object.h#oldcode3631
runtime/vm/object.h:3631: ASSERT(sizeof(RawCode) == OFFSET_OF(RawCode, data_));
On 2014/06/04 15:53:14, Ivan Posva wrote:
> There are too many things riding on the assertion that sizeof(RawCode) gives
us
> the correct information. We will have to solve this differently.

Yes, this is fragile: InstanceSize(intptr_t len) needs to be correct. This means
that RoundedAllocationSize must match what the C++ compiler produces. There are
differences betwwen MSVC and g++: sizeof(RawCode) is 72 on my Linux machine, and
80 on Win32.

OFFSET_OF(data_) is 72 on Linux, and 76 on Win32.

In this instance the RoundedAllocationSize would be still 80 since we round up
to multiples of 8 (kObjectAlignment), but there is no guarantee that these
things match.