[rfc][icedtea-web] PR1592 reproducers update

[Jiri Vanek]

On 01/20/2014 09:49 PM, Andrew Azores wrote:
> On 01/20/2014 11:07 AM, Jiri Vanek wrote:
>> On 01/09/2014 08:06 PM, Andrew Azores wrote:
>>> On 01/09/2014 11:17 AM, Andrew Azores wrote:
>>>> On 01/03/2014 02:43 PM, Andrew Azores wrote:
>>>>> Updated PR1592 tests, using a custom reproducer rather than split simple/signed. This allows
>>>>> method calls to be made in the normal way as well as via reflection. JNLP includes both
>>>>> applications and applets now, and they close properly as well.
>>>>>
>>>>> (snip)
>>>>>
>>>>> Thanks,
>>>>> Andrew A
>>>>
>>>> Went back over this and realized one of the tests was written wrong. The
>>>> assertAccessControlException helper method in the testcase file is now a little stricter about the
>>>> type of AccessControlException (so that the exceptions due to applets not being allowed to call
>>>> System.exit don't falsely fulfill this assertion), and MixedSigningAppletHelper.attackDoPrivileged
>>>> now properly calls MixedSigningAppletSigned#testSignedReadPropertiesDoPrivileged, as it should
>>>> have been doing. In this case, the Unsigned JAR actually *is* meant to be able to retrieve data
>>>> from the Signed JAR (as is the point of the AccessController.doPrivileged call), so the testcases
>>>> now expect this test to successfully read from System.getProperty, rather than receive an
>>>> AccessControlException. However, the tests still verify that in situations where the Signed JAR
>>>> has a method call that involves a privileged action *without* being placed inside a doPrivileged
>>>> call, an AccessControlException will be thrown if the Unsigned code attempts to access it, as
>>>> expected.
>>>>
>>>> Thanks,
>>>>
>>>
>>> Sorry, please ignore the previous patch. The extra changes were not made based on the most recent
>>> other changes. Attached are the properly rebased patches, also split into three as they were
>>> originally.
>>>
>>> Thanks,
>>>
>>
>> Thanx for ping.
>> There is really many of jnlps which are nearly similar. Maybe better idea can be to have one template, and generate all the rest from it?
>>
>> I altready did this - and generated them ion BeforeClass.
>>
>> What do you think?
>
> I like this idea, but I didn't know we were okay with having reproducers do tricks like this ;)
>
> Thanks,
>


On seriosu flaw: you have "private static final ServerAccess server = new ServerAccess();" declard. by this you are owerwritting the one in BrowserTest, so no browser test will work, and wil fial 'can not lunch unset browser".
Just remove this line.

Also - non of the jnlp have security element specified. It i s intentional?? I thought it was an reason for this test to be redone.

After fixed first, and explined second, ok to head.

J.