Mojo: Small fixes/cleanup to MessageInTransit, before I factor out the secondary buffer.

Mojo: Small fixes/cleanup to MessageInTransit, before I factor out the secondary buffer.

I'm making these changes as I write the factored-out secondary buffer in
parallel, but it's probably to review these changes now rather than as a
part of the move/refactor.

Notable changes:
* Use the actual size for the secondary buffer, rather than the
  estimated size. (Probably this makes no difference now.)
* Discard the dispatchers earlier (in SerializeAndCloseDispatchers()).

R=darin@chromium.org

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=267542

[darin (slow to review), 2014-05-01 04:13:50]

LGTM

https://codereview.chromium.org/260823002/diff/1/mojo/system/message_in_trans...
File mojo/system/message_in_transit.cc (right):

https://codereview.chromium.org/260823002/diff/1/mojo/system/message_in_trans...
mojo/system/message_in_transit.cc:311: // There's no aligned realloc, so it's no
good way to release unused space (if
would it be worth trying to invent an aligned realloc? is this a TODO?

[viettrungluu, 2014-05-01 16:54:20]

Thanks.

On 2014/05/01 04:13:50, darin wrote:
> LGTM
> 
>
https://codereview.chromium.org/260823002/diff/1/mojo/system/message_in_trans...
> File mojo/system/message_in_transit.cc (right):
> 
>
https://codereview.chromium.org/260823002/diff/1/mojo/system/message_in_trans...
> mojo/system/message_in_transit.cc:311: // There's no aligned realloc, so it's
no
> good way to release unused space (if
> would it be worth trying to invent an aligned realloc? is this a TODO?

Arguably a TODO (i.e., I thought about it)? It's not a big deal right now, so
I'm not too worried about it.

Unfortunately, a brief survey of various memory allocation APIs seems to
indicate that they never seem to have an aligned realloc. So it's something that
we'd have to, e.g., add to tcmalloc.

OTOH, our alignment requirements are so minimal that quite possibly a regular
malloc() would always yield suitably-aligned memory anyway....

[viettrungluu, 2014-05-01 16:56:11]

Committed patchset #1 manually as r267542 (presubmit successful).