FBTF: Reapplies r55259, the first new IPC definition patch.

FBTF: Reapplies r55259, the first new IPC definition patch.

This moves MessageWithTuple::Read() back into the main ipc_message_utils.h
header from the private ipc_messsage_utils_impl.h header. In release mode, this
was causing link failures.

BUG=51411
TEST=none

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=55587

[Elliot Glaysher, 2010-08-09 20:30:04]

The patch is the same as before, except that the definition of Read() has been
moved back inline. I've confirmed that this compiles in release mode locally
(while it didn't before making this change).

[Matt Perry, 2010-08-09 20:56:28]

LGTM, though I only looked at ipc_message_utils*

http://codereview.chromium.org/3069034/diff/1/24
File ipc/ipc_message_utils.h (right):

http://codereview.chromium.org/3069034/diff/1/24#newcode1038
ipc/ipc_message_utils.h:1038: static bool Read(const Message* msg, Param* p) {
Interesting, it still compiles with the forward declarations when this method is
defined here?

[Elliot Glaysher, 2010-08-09 21:04:03]

http://codereview.chromium.org/3069034/diff/1/24
File ipc/ipc_message_utils.h (right):

http://codereview.chromium.org/3069034/diff/1/24#newcode1038
ipc/ipc_message_utils.h:1038: static bool Read(const Message* msg, Param* p) {
On 2010/08/09 20:56:28, Matt Perry wrote:
> Interesting, it still compiles with the forward declarations when this method
is
> defined here?

Correct. Since it is a reference type (and we don't dereference it), the
compiler doesn't need to know the size of Param, only that the type exists and
that ReadParam<Param>() exists.