SPDY: Make sure we don't try to send https/wss over an unauthenticated, but encrypted SSL socket.

SPDY: Make sure we don't try to send https/wss over an unauthenticated, but encrypted SSL socket.
BUG=46924
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=50997

[willchan no longer on Chromium, 2010-06-27 18:28:19]

I need to write a test for this, but I'd like to send this out for review first
and land it in time for the dev channel if possible.

[Mike Belshe, 2010-06-27 21:15:40]

Thanks for doing this, Will.

Let me know what you think of erroring the stream altogether.

http://codereview.chromium.org/2805039/diff/1/4
File net/spdy/spdy_session.cc (right):

http://codereview.chromium.org/2805039/diff/1/4#newcode281
net/spdy/spdy_session.cc:281: DCHECK(!is_secure_ ||
I think we should fail it - a DCHECK is not strong enough.  We can't send https
urls to a unauthenticated server, ever...  This function doesn't have a good
mechanism for erroring, though.

[willchan no longer on Chromium, 2010-06-28 15:27:27]

Sorry, slip of the mind.  I blame it on the Boston weather.  I've addressed this
now by storing the certificate error and returning it whenever trying to request
a secure URL over an unauthenticated session.

On 2010/06/27 21:15:40, Mike Belshe wrote:
> Thanks for doing this, Will.
> 
> Let me know what you think of erroring the stream altogether.
> 
> http://codereview.chromium.org/2805039/diff/1/4
> File net/spdy/spdy_session.cc (right):
> 
> http://codereview.chromium.org/2805039/diff/1/4#newcode281
> net/spdy/spdy_session.cc:281: DCHECK(!is_secure_ ||
> I think we should fail it - a DCHECK is not strong enough.  We can't send
https
> urls to a unauthenticated server, ever...  This function doesn't have a good
> mechanism for erroring, though.

[Mike Belshe, 2010-06-28 15:37:52]

LGTM

Any way to build a test case for this?

[willchan no longer on Chromium, 2010-06-28 17:30:06]

As discussed with Mike, I'm landing first and adding the test later.

On 2010/06/28 15:37:52, Mike Belshe wrote:
> LGTM
> 
> Any way to build a test case for this?