test failure security/krb5/auto/SpnegoReqFlags.java [was - Re: OpenJDK6 B24]

[Weijun Wang]

 > FWIW, backporting the framework is not adequate. Though the test
 > compiles after that, it still fails saying:

Yes, it's tough.

The basic test framework is only Action+Context+KDC+OneKDC. However, 
every now and then, with each new bug/RFE in Kerberos, these classes are 
enhanced to simulate new functions as well as new errors. Since not all 
code changes in JDK 7 is backported into JDK 6, picking out changes in 
the framework is quite complicated.

Now I guess the best we can do is to remove the test and pray that if it 
runs fine in JDK 7 (and now in 8) it should also run fine in 6. Also, 
our sustaining team did backport all necessary tests into Oracle's JDK 6 
update releases. Hopefully that also guarantees something.

-Max


On 11/10/2011 12:20 PM, Deepak Bhole wrote:
> * Lana Steuck<lana.steuck at oracle.com>  [2011-11-09 22:40]:
>> Below is an update about OpenJdk6 b24 release.
>>
>> There were a number of minor issues with the build (infrastructure,
>> not related to jdk6 code) which got resolved.
>> I had built openjdk6 on all 8 platforms. There is a new test failure
>> on linux_i586 (and all other platforms).
>>
>> TEST: sun/security/krb5/auto/SpnegoReqFlags.java
>> TEST RESULT: Failed. Compilation failed: Compilation failed
>> (detailed error message is below).
>>
>> By looking at the openjdk archive, it seemed that the test was
>> backported from 7 on Sep 6, 2011.
>> Email Request: http://mail.openjdk.java.net/pipermail/jdk6-dev/2011-September/002506.html
>> Hg notification: http://mail.openjdk.java.net/pipermail/jdk6-dev/2011-September/002511.html
>>
>> Weijun Wang from JDK Security team looked at the failure (thank you,
>> Weijun!), below is his explanation:
>>
>
> Doh. Sorry about this. I had run into this as well and was in the middle
> of investigating the issue when it slipped through my list thanks to our
> hectic office move recently :(
>
> FWIW, backporting the framework is not adequate. Though the test
> compiles after that, it still fails saying:
>
> java.io.IOException: DerInputStream.getLength(): lengthTag=127, too big.
>
>
> Brock Janiczak from Red Hat helped me out on that one and determined
> that it was an additional bug whereby pdata was being regarded as
> mandatory when it's optional:
>
> http://hg.openjdk.java.net/jdk7/tl/jdk/rev/7360321c37e3
>
> After that patch, the tests get further, but one still fails saying:
>
> RABBIT.HOLE>  dummy sends AS-REQ for krbtgt/RABBIT.HOLE
> KrbException: Additional pre-authentication required (25)
>
> That was as far as I got before I lost track :/
>
> The failures all seem to be due to unrelated bugs however and I can
> confirm that the ported patches do indeed fix the specific issues we
> were seeing.
>
> Sorry again for the slippage.
>
> Cheers,
> Deepak
>