Patch: issue RT-14177

[Kevin Rushforth]

 > I do wonder whether we are being consistent in the other places where 
classloading comes into play.

I know that was David's concern when he initially commented on it and 
reassigned it. In addition to other possible places in controls, has 
this been rationalized with the recent changes in behavior to FXML loading?

Also, as a follow-on to the security concern I raised: given that it 
can't / shouldn't be run in a doPrivileged block, you will need to make 
sure that it doesn't blow up in the normal use case when run as an 
applet (as long as the original code is executed first and the new code 
is only done as a fallback in case that fails it should be fine).

-- Kevin


Jonathan Giles wrote:
> +1 on the patch, although I do wonder whether we are being consistent 
> in the other places where classloading comes into play. Is this 
> something that should be extracted and used as an internal library?
>
> For example, there is much the same code in PopupControl that needs to 
> be fixed at the same time as the patch you are submitting here, but 
> more abstractly, should we have a library rather than duplicate 
> similar code in FXML and CSS as well?
>
> -- Jonathan
>
>
> On 1/06/2012 3:14 p.m., Richard Bair wrote:
>> Florian / Tom,
>>
>> I've uploaded two patches to the issue:
>> rt-14177.2.patch
>> rt-14177.3.patch
>>
>> The .3 patch is my preferred one. Can you take a look / apply it and 
>> see if it solves your issue? If the .3 patch works, then I think 
>> we've got a clean patch which will produce no additional performance 
>> overhead to a correctly functioning JavaFX application today, but 
>> will fix your issue (at least as Tom said in the issue itself for 95% 
>> of the folks).
>>
>> We can then file a new issue for the remaining 5% cases which will 
>> need more time (since it involves some form of new API) whereas in 
>> this case it should "just work".
>>
>> David / Jonathan, can you guys give a quick code review? I'm pretty 
>> sure the code does what I want it to do, and hopefully Florian / Tom 
>> will tell us if what I want it to do is actually the right thing to 
>> do (ie: solves the issue) :-).
>>
>> Thanks
>> Richard
>>
>> On May 31, 2012, at 5:45 PM, Richard Bair wrote:
>>
>>> Hi Florian,
>>>
>>> We'll get this fixed ASAP.
>>>
>>> Thanks
>>> Richard
>>>
>>> On May 31, 2012, at 4:45 PM, Kevin Rushforth wrote:
>>>
>>>> Hi Florian,
>>>>
>>>> This JIRA issue is targeted for 3.0 which is just in the initial 
>>>> planning stages, so no one has looked at it yet.
>>>>
>>>> The patch as attached to the issue deals with one case of CSS 
>>>> loading a skin, but when the controls team evaluated it, their take 
>>>> was that it a more general platform issue, although it could be 
>>>> dealt with in isolation. In any case, even if I were to assign this 
>>>> back to someone on the controls team, this code would need to be 
>>>> evaluated for potential bugs and to ensure that there are no 
>>>> security issues (probably not an issue, but since it would be using 
>>>> a system class loader we need to make sure it is never called from 
>>>> a privileged context), which is unlikely to happen for the current 
>>>> release.
>>>>
>>>> -- Kevin
>>>>
>>>>
>>>> Florian Brunner wrote:
>>>>> Hi Kevin,
>>>>>
>>>>> What is the status of my patch?
>>>>> As mentioned in JIRA, this issue is becoming a blocker! Please 
>>>>> increase the priority! (Nobody from Oracle is responding on JIRA).
>>>>>
>>>>> Can we agree then to integrate the provided patch? It improves the 
>>>>> situation but doesn't introduce any API additions and we can 
>>>>> replace/ complete it later with other solutions if needed.
>>>>> Please tell me if I should change something in the patch to get it 
>>>>> approved.
>>>>>
>>>>> Regards,
>>>>> Florian
>>>>>
>>>>> Am Freitag 06 April 2012, 23:58:13 schrieb Kevin Rushforth:
>>>>>
>>>>>> I attached it to the JIRA issue.
>>>>>>
>>>>>> -- Kevin
>>>>>>
>>>>>>
>>>>>> Florian Brunner wrote:
>>>>>>
>>>>>>> Here the patch created with "hg diff --git" and zipped.
>>>>>>>
>>>>>>> Regards,
>>>>>>> Florian
>>>>>>>
>>>>>>> Am Freitag 06 April 2012, 18:31:23 schrieb Kevin Rushforth:
>>>>>>>
>>>>>>>> Right. Since you didn't file the issue and are not the 
>>>>>>>> assignee, you can't add an attachment.
>>>>>>>>
>>>>>>>> Copying Brian Beck in case this is something we can (and want 
>>>>>>>> to) fix.
>>>>>>>>
>>>>>>>> In the mean time, you can e-mail me the patch and I'll attach 
>>>>>>>> it to the bug report.
>>>>>>>>
>>>>>>>> -- Kevin
>>>>>>>>
>>>>>>>>
>>>>>>>> Florian Brunner wrote:
>>>>>>>>
>>>>>>>>> Hi Kevin,
>>>>>>>>>
>>>>>>>>> Thanks for your response.
>>>>>>>>>
>>>>>>>>> Unfortunatly, I don't see a way to attach something to the 
>>>>>>>>> JIRA issue.
>>>>>>>>>
>>>>>>>>> Either I'm missing something or maybe I need additional rights?
>>>>>>>>>
>>>>>>>>> Regards,
>>>>>>>>> Florian
>>>>>>>>>
>>>>>>>>> Am Donnerstag 05 April 2012, 18:54:38 schrieben Sie:
>>>>>>>>>
>>>>>>>>>> Great. Please attach the patch as a zip file to the JIRA 
>>>>>>>>>> issue (preferably generated with either webrev or "hg diff 
>>>>>>>>>> --git") so that it can be evaluated.
>>>>>>>>>>
>>>>>>>>>> -- Kevin
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> Florian Brunner wrote:
>>>>>>>>>>
>>>>>>>>>>> Hi,
>>>>>>>>>>>
>>>>>>>>>>> Let me introduce myself: my name is Florian Brunner and I'm 
>>>>>>>>>>> a Senior Software Engineer from Switzerland.
>>>>>>>>>>>
>>>>>>>>>>> I started to use JavaFX in an OSGi environment. Now I 
>>>>>>>>>>> stumbled over the following issue, which blocks me:
>>>>>>>>>>> http://javafx-jira.kenai.com/browse/RT-14177
>>>>>>>>>>>
>>>>>>>>>>> I have written a patch now. It probably doesn't solve the 
>>>>>>>>>>> whole issue, but at least it unblocks me for now.
>>>>>>>>>>>
>>>>>>>>>>> It would be great if you could integrate it.
>>>>>>>>>>>
>>>>>>>>>>> Note that I contributed to OpenJDK before (Swing) and have 
>>>>>>>>>>> signed the Sun Contributor Agreement:
>>>>>>>>>>>
>>>>>>>>>>> http://www.oracle.com/technetwork/community/oca-486395.html#b
>>>>>>>>>>> Florian Brunner - java.net - puce
>>>>>>>>>>>
>>>>>>>>>>> Regards,
>>>>>>>>>>> Florian
>>>>>>>>>>>
>>>>>