One new change (was Re: 2nd round code review request: 7055363: jdk_security3 cleanup)

Weijun Wang weijun.wang at oracle.com
Sun Aug 7 03:19:35 UTC 2011



On 08/04/2011 10:38 AM, Xuelei Fan wrote:
> On 8/4/2011 10:34 AM, Weijun Wang wrote:
>> Oh, that's true.
>>
>> I'll add it back. Why did I remove it at the first place?
>>
> You may not have to add it back, the library tag may help to build the
> dependent classes. Maybe, we have some incorrect understand of the
> "build" tag.

A @build is needed here. The ResetModule file is loaded as a JAAS 
configuration provider. The class is not directly referenced in the test 
and will not be automatically compiled.

In fact, JAAS uses the thread context class loader to load a 
Configuration provider, so even if ResetModule is built, it still cannot 
be loaded.

I've update the test with a dirty hack:

--- a/test/javax/security/auth/login/LoginContext/ResetConfigModule.java
+++ b/test/javax/security/auth/login/LoginContext/ResetConfigModule.java
@@ -25,7 +25,6 @@
   * @test
   * @bug 4633622
   * @summary  bug in LoginContext when Configuration is subclassed
- *
   * @build ResetConfigModule ResetModule
   * @run main ResetConfigModule
   */
@@ -40,32 +39,40 @@

      public static void main(String[] args) throws Exception {

+        Thread.currentThread().setContextClassLoader(
+                ResetConfigModule.class.getClassLoader());

Please confirm this is OK. Or, do you have a better solution?

BTW, I've asked about this in another mail but no reply yet:

http://mail.openjdk.java.net/pipermail/core-libs-dev/2011-August/007286.html

Thanks
Max

>
> Xuelei
>
>> Thanks
>> Max
>>
>> On 08/04/2011 10:11 AM, Xuelei Fan wrote:
>>> . test/javax/security/auth/login/LoginContext/ResetConfigModule.java
>>>
>>> You removed the "run" tag, and "build" tag is the last tag. It seems
>>> that the test will only run building, but not run the class.main(). I
>>> got the strange behavior when I run some tests with build tag with jtreg
>>> 4.1 b03. Did you come into the same issue?
>>>
>>> Thanks,
>>> Xuelei
>>>
>>> On 8/3/2011 11:50 AM, Weijun Wang wrote:
>>>>
>>>>
>>>> On 08/03/2011 12:43 AM, Alan Bateman wrote:
>>>>> Weijun Wang wrote:
>>>>>> Hi All
>>>>>>
>>>>>> http://cr.openjdk.java.net/~weijun/7055363/webrev.01/
>>>>> I went through the changes to the tests and they look okay to me. It's
>>>>> good to see so many tests removed from the problem list and the tests
>>>>> updated to work in samevm or agentvm mode.
>>>>>
>>>>> Is there a bug created for the ec tests that are failing on
>>>>> solaris-i586? If so then it would be good to include the bugID for
>>>>
>>>> I thought there was one but I cannot find it. *Vinnie*, do you know it?
>>>>
>>>>> future edits to the problem list. Also I see you've commented out
>>>>> two ec
>>>>> tests and I wonder if these lines should be removed or uncommented.
>>>>
>>>> I don't understand why these 2 pass now. Maybe they still fail on some
>>>> platforms not in JPRT? Anyway, I just don't want to purge the lines.
>>>>
>>>>>
>>>>> test/com/sun/security/auth/login/ConfigFile/IllegalURL.java - no
>>>>> problem
>>>>> with your change but I notice this test using URL's getFile which isn't
>>>>> the right way to get to a file path. Could this be changed to using
>>>>> File(URI)?
>>>>
>>>> This test is a compatibility check, see
>>>>
>>>>      http://hg.openjdk.java.net/jdk7/jsn/jdk/rev/4e7ad09de58b
>>>>
>>>> I remember there were some strange things on Windows, say, UNC, \ vs /,
>>>> that made some hand-crafted URL illegal. We cannot support all of them,
>>>> so just make sure the old supported ones are still supported.
>>>>
>>>> Thanks
>>>> Max
>>>>
>>>>>
>>>>> -Alan.
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>
>



More information about the security-dev mailing list