In rare cases (when running inside QEMU), the event...

In rare cases (when running inside QEMU), the event
object is not yet in the signaled state after a
Winsock function succeeds synchronously, so it is
necessary to wait for it to become signaled.

R=eroman
BUG=6500
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=8989

[wtc, 2009-01-30 20:37:24]

eroman: please review this CL.

darin: just FYI.  QEMU testing shows that in rare cases,
the event object may not be signaled yet after a Winsock
function succeeds synchronously.  So we still need to
wait for it to become signaled.  Since in most cases
the event object is already signaled at that point, I just
do the wait on the IO thread to avoid a context switch.

[darin (slow to review), 2009-01-30 21:00:54]

LGTM

[eroman, 2009-01-30 21:44:46]

lgtm

http://codereview.chromium.org/19515/diff/9/10
File net/base/tcp_client_socket.h (right):

http://codereview.chromium.org/19515/diff/9/10#newcode89
Line 89: // After a Winsock function succeeds synchronously, waits for the
This should probably be wrapped in a:

#if defined (OS_WIN)
...
#endif

Unless it will be implemented on other platforms.

[wtc, 2009-01-30 21:48:06]

http://codereview.chromium.org/19515/diff/9/10
File net/base/tcp_client_socket.h (right):

http://codereview.chromium.org/19515/diff/9/10#newcode89
Line 89: // After a Winsock function succeeds synchronously, waits for the
On 2009/01/30 21:44:46, eroman wrote:
> This should probably be wrapped in a:
> 
> #if defined (OS_WIN)
> ...
> #endif

Yes, this is a Windows-specific method.  It is already wrapped in a #if
defined(OS_WIN).

[eroman, 2009-01-30 21:55:17]

http://codereview.chromium.org/19515/diff/9/10
File net/base/tcp_client_socket.h (right):

http://codereview.chromium.org/19515/diff/9/10#newcode89
Line 89: // After a Winsock function succeeds synchronously, waits for the
On 2009/01/30 21:48:06, wtc wrote:
> On 2009/01/30 21:44:46, eroman wrote:
> > This should probably be wrapped in a:
> > 
> > #if defined (OS_WIN)
> > ...
> > #endif
> 
> Yes, this is a Windows-specific method.  It is already wrapped in a #if
> defined(OS_WIN).
> 
> 

Oh right, didn't notice!

LGTM.