Code review request: 8017453: ReplayCache tests fail on multiple platforms (was Re: Please be prepared for possible test failures (was Fwd: [JBS] (JDK-8001326) Improve Kerberos replay caching))

[Weijun Wang]

On 6/24/13 9:49 AM, Xuelei Fan wrote:
> ReplayCacheTestProc.java
> ========================
> 75       String os = System.getProperty("os.name", "???");
> 
> What happens if "os" is "???"?

Then it's neither Windows nor Mac OS X. :)

I gave a default value so that the result will not be null. Although the
system property "os.name" should always have a value, it makes me
comfort not thinking of possible NPE.

Thanks
Max

> 
> Xuelei
> 
> On 6/24/2013 12:36 AM, Wang Weijun wrote:
>> Send again. And BTW, JPRT runs fine.
>>
>> 在 Jun 23, 2013，6:29 PM，Weijun Wang <weijun.wang at oracle.com> 写道：
>>
>>> The macosx problem found, the machine's native GSS does not support shared replaycache.
>>>
>>> *Valerie* and/or *Xuelei*, can you please review the fix?
>>>
>>>    http://cr.openjdk.java.net/~weijun/8017453/webrev.00/
>>>
>>> So interop (between Java and native) will not run on Windows and Mac now. It still tests shared replaycache between Java processes.
>>>
>>> The left problem is linux-i586 test running on a x64 machine. Either we can install i386 libraries (ia32-libs for Ubuntu) or disallow 32 bit tests on 64 bit systems.
>>>
>>> I'm running JPRT now.
>>>
>>> Thanks
>>> Max
>>>
>>>
>>> On 6/22/2013 11:17 PM, Weijun Wang wrote:
>>>>
>>>>
>>>> On 6/22/13 4:48 PM, Weijun Wang wrote:
>>>>> I've found some reasons. There are several kinds of failures:
>>>>>
>>>>> 1. macosx. Reason not known yet.
>>>>>
>>>>> 2. windows. Tests should not run at all.
>>>>
>>>> Will not call native GSS on Windows.
>>>>
>>>>>
>>>>> 3. linux-i586. Even an existing native GSS test fails. This is probably
>>>>> the case we need to update the system. It's also possible the earlier
>>>>> rcache failure triggers this one, but not likely.
>>>>
>>>> The machine is a x86_64 and it has
>>>> /usr/lib/x86_64-linux-gnu/libgssapi_krb5.so.2. But the test is linux-i586.
>>>>
>>>> Maybe we should only test linux-i586 on x86 machines?
>>>>
>>>>>
>>>>> 4. solaris-sparc. The failure is not the interop test, but a pure Java
>>>>> one (also added in this changeset). Should be easy to evaluate.
>>>>
>>>> That machine has a /etc/krb5.conf which sets "clockskew = 3600", quite
>>>> huge value. I'll add a krb5.conf to the test to override it.
>>>>
>>>> So the last one to check is the macosx one. If I cannot solve the
>>>> problem on Sunday, will problem-list it.
>>>>
>>>> Thanks
>>>> Max
>>>>
>>>>>
>>>>> I made a final code update (due to CCC update, change environment
>>>>> variable to system property) and haven't run JPRT after that. I thought
>>>>> it's small. At least #2 above is because of this. :-(
>>>>>
>>>>> Thanks
>>>>> Max
>>>>>
>