[8] Request for Review: 8007967: Infinite loop can happen in sun.security.provider.certpath.SunCertPathBuilder.depthFirstSearchForward()

[Jason Uh]

Thanks, Sean.

I've updated the webrev with your suggestions. Also, I've changed the 
check for loops in ForwardBuilder.java to bail out whenever the same 
certificate is encountered twice.

Please note that none of the policy mapping tests in the NIST PKITS use 
loops. Also, RFC 4158's section on Certification Path building states 
that X.509 "requires that certificates are not repeated when building 
paths," but does not mention policy mapping. It should therefore be okay 
to make this change.

Please take a look:
http://cr.openjdk.java.net/~juh/8007967/webrev.01/

Thanks,
Jason

On 12/04/2013 01:18 PM, Sean Mullan wrote:
> Just 2 comments on DistributionPointFetcher:
>
> You can eliminate some duplication of code by changing the existing
> getCRLs method to just call the new method with a null prevCert parameter.
>
> On lines 659-663, you don't need to add @code tags, that is only for
> javadoc comments.
>
> --Sean
>
> On 12/03/2013 01:51 PM, Jason Uh wrote:
>> Could I please get a review for this change? This change fixes some
>> issues in CertPath building and CRL verification. The main components of
>> this fix are:
>>
>> 1. Proper setting of TrustAnchors when verifying indirect CRLs obtained
>> from CRL Distribution Points. I added an overloaded getCRLs() method to
>> DistributionPointFetcher for this.
>>
>> 2. Terminating the CertPath build immediately when the target cert is
>> found to be revoked.
>>
>> 3. Some clarification in the comments.
>>
>> Webrev: http://cr.openjdk.java.net/~juh/8007967/webrev.00/
>> Bug: https://bugs.openjdk.java.net/browse/JDK-8007967
>>
>> Thanks,
>> Jason
>