david.grieve at oracle.com
Mon Oct 22 08:56:05 PDT 2012
The bug was cloned from another and that patch is the one for the larger leak and is in 2.2 already. There is no plan for a patch for this issue in any 2.2.x release, as far as I know.
If you can use existing pseudo class (selected, disabled, focused, etc.) state in your styles, rather than style classes, that will help.
On Oct 22, 2012, at 11:39 AM, Scott Palmer wrote:
> Thanks guys. RT-23691 might be related to what we are observing. We aren't even 100% convinced that FX is leaking. I need to study it a bit more. Attempts at creating a test case that reproduces our issue have not panned out yet.
> I see that there is a patch attached to that bug. What is the state of the patch? Is it going to be applied to 2.2.6? Sound like it is already too late for 2.2.4, but even that is probably too late for my release anyway, so I'm more interested in workarounds at this point.
> Is there an easy way to get a list of issues resolved between releases?
> On 2012-10-22, at 10:06 AM, David Grieve <david.grieve at oracle.com> wrote:
>> There was a larger leak discovered late in the 2.2 release cycle. This leak was fixed but another, smaller, leak remained. Unfortunately, this smaller leak was not fixed because there simply wasn't enough time left in the release. The issue is being tracked by bug RT-23691. The issue is not yet resolved.
>> On Oct 20, 2012, at 6:01 PM, Scott Palmer wrote:
>>> Thanks for the response. 8.0 is too far away. I need to ship my app in a
>>> few weeks. 2.2.3 may be old to you guys, but I just found 2.2.3
>>> yesterday. We were on 7u7 and then I discovered that there was a new
>>> JavaFX runtime in 7u9.. Would have been nice to know! 7u7 shipped not too
>>> long ago with 2.2.0 and I didn't see any sort of announcements that there
>>> was anything newer.
>>> I will try to isolate the leak. We think we are seeing
>>> com.sun.javafx.css.StyleHelper$CacheEntry and
>>> com.sun.javafx.css.StyleHelper$CalculatedValue leaking (at least).
>>> jvisualvm shows the number of those objects are always growing. Comparing
>>> heap dumps we have never observed the number to decrease.
>>> I'm currently only able to connect to the machine running my test via
>>> Remote Desktop, and the screen saver probably keep kicking in, I think that
>>> is affecting the test. Is JavaFX smart enough to not render when the
>>> screen is blanked?
>>> I will try to produce a test case. But if anyone has knowledge about this
>>> that wants to share, any info is appreciated.
>>> On Sat, Oct 20, 2012 at 5:16 PM, Jonathan Giles
>>> <jonathan.giles at oracle.com>wrote:
>>>> David Grieve is the go-to expert on all things CSS, but from a
>>>> higher-level perspective the general answer is always: you tell us! :-)
>>>> There might be a leak we don't know about, or can't easily reproduce, and
>>>> you might have just the information we need to finally get rid of that
>>>> annoying leak you mention.
>>>> So, please, file Jira bugs on us and we'll get to looking at them as soon
>>>> as possible. If you can provide a reproducible test case that would be
>>>> hugely appreciated too.
>>>> Finally, from our point of view JavaFX 2.2.3 is rather old code - we've
>>>> been actively developing in the 8.0 branch for quite a long time now, and
>>>> one of our primary focuses during this time has been performance (CPU and
>>>> memory). In other words, there is a chance it may already be resolved (or
>>>> minimised) in the 8.0 branch. I know it is a lot to ask, but there are 8.0
>>>> builds available that you can test with. If you have the time and
>>>> inclination, it might be worth testing against these builds to see if it is
>>>> -- Jonathan
>>>> On 21/10/2012 2:10 a.m., Scott Palmer wrote:
>>>>> Are there known leaks in 2.2.3 with regards to CSS and style cache stuff?
>>>>> In my application I add and remove style classes to highlight objects in
>>>>> the scene graph that are in various states (selected, active, disabled,
>>>>> etc). It seems that every time the style class list is changed that there
>>>>> is a small leak. Since while the app runs and process data (a process that
>>>>> may run indefinitely) the state of objects in the graph changes as data is
>>>>> processed, this is a significant issue. Our customers may have the app
>>>>> fail after several hours or a few days because of these leaks.
>> David Grieve | Principal Member of Technical Staff
>> Mobile: +16033121013
>> Oracle Java Client UI and Tools
>> Durham, NH 03824
>> <green-for-email-sig_0.gif> Oracle is committed to developing practices and products that help protect the environment
David Grieve | Principal Member of Technical Staff
Oracle Java Client UI and Tools
Durham, NH 03824
Oracle is committed to developing practices and products that help protect the environment
More information about the openjfx-dev