On Tue, 2012-12-11 at 11:52 -0800, Juan Lang wrote:
On Tue, Dec 11, 2012 at 6:10 AM, Hans Leidekker hans@codeweavers.com wrote: On Tue, 2012-12-11 at 14:52 +0100, Jacek Caban wrote: > On 12/11/12 09:45, Hans Leidekker wrote: > > https://testbot.winehq.org/JobDetails.pl?Key=23300 is a test which shows that > > revocation checks fail for the certificate on outlook.com when passed straight > > to CertVerifyRevocation. The reason is that a CRL link specified in the > > certificate does not resolve. > > > > https://testbot.winehq.org/JobDetails.pl?Key=23301 is a test which makes > > a secure connection to outlook.com from wininet and shows that this succeeds. > > > > My conclusion is that native wininet doesn't perform revocation checks. > > Your tests prove that we should relax our verification on > CERT_TRUST_IS_OFFLINE_REVOCATION or something similar. To prove that > revocation checks are not made, a test with truly revoked cert would be > needed.
True, though to perform the revocation check the CRL has to be retrieved and my tests with wireshark didn't show any signs of that.Would adding to the tests as part of this patch be a bad thing?
I thought about that but I am hesitant to use a random site that's not under our control.
The alternative is to setup our own site with a certificate that only fails the revocation check, which I think means that we need to have it signed by a trusted root and then revoked. I'm not sure we have the means to do that currently.
Do you have any suggestions?