Properly let the SSL engine know when the underlying stream is closed....

Properly let the SSL engine know when the underlying stream is closed.

BUG=http://crbug.com/14196
TEST=NONE (this fix is related to the bug but does not fix it)


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=20864

[Avi (use Gerrit), 2009-07-15 22:18:36]

Streams returning 0 mean end-of-stream. Dunno why I didn't handle this before...

[Avi (use Gerrit), 2009-07-15 22:31:36]

Rohit's around; he's taking it.

[John Grabowski, 2009-07-15 22:39:54]

LGTM

http://codereview.chromium.org/155595/diff/1006/1007
File net/socket/ssl_client_socket_mac.cc (right):

http://codereview.chromium.org/155595/diff/1006/1007#newcode719
Line 719: else if (rv == 0) // stream closed
Shouldn't this be (rv == OK) to be consistent with constant used in
NetErrorFromOSStatus() for conversion from errSSLClosedGraceful?

[Avi (use Gerrit), 2009-07-15 22:47:08]

http://codereview.chromium.org/155595/diff/1006/1007
File net/socket/ssl_client_socket_mac.cc (right):

http://codereview.chromium.org/155595/diff/1006/1007#newcode719
Line 719: else if (rv == 0) // stream closed
On 2009/07/15 22:39:54, John Grabowski wrote:
> Shouldn't this be (rv == OK) to be consistent with constant used in
> NetErrorFromOSStatus() for conversion from errSSLClosedGraceful?

The docs say "zero" is returned for end-of-stream (see socket.h) and for balance
(we're doing <0) I'm sticking to 0.

[wtc, 2009-07-16 00:22:50]

http://codereview.chromium.org/155595/diff/1006/1007
File net/socket/ssl_client_socket_mac.cc (right):

http://codereview.chromium.org/155595/diff/1006/1007#newcode719
Line 719: else if (rv == 0) // stream closed
On 2009/07/15 22:47:08, Avi wrote:
>
> The docs say "zero" is returned for end-of-stream (see socket.h) and for
balance
> (we're doing <0) I'm sticking to 0.

Yes, Read() returns a byte count or error code, so 0 is
approriate here.