<AWT Dev> sun.awt.X11.XEvent is creating 600 MB of char[] for no good reason

Federico federicotg at gmail.com
Wed Oct 26 12:27:22 PDT 2011


Almost all sun.awt.X11.XWrapperBase subclasses have similarly horrible
getFieldsAsString() implementations, but they are all automatically
generated (which is a relief as there was no human being actually
writing such code).

They all say at the beginning:

// This file is an automatically generated file, please do not edit
this file, modify the WrapperGenerator.java file instead !

My guess is in file
/openjdk/jdk/src/solaris/classes/sun/awt/X11/generator/WrapperGenerator.java

in line 678
public void writeToString(StructType stp, PrintWriter pw) {

starts the code generating bit.


2011/10/26 Oleg Sukhodolsky <son.two at gmail.com>:
> +1 to Artem.
>
> It is better to fix this too.
>
> Oleg.
>
> On Wed, Oct 26, 2011 at 7:34 PM, Artem Ananiev <artem.ananiev at oracle.com> wrote:
>>
>> On 10/26/2011 7:26 PM, Anthony Petrov wrote:
>>>
>>> On 10/26/2011 7:12 PM, Artem Ananiev wrote:
>>>>
>>>> On 10/21/2011 4:00 PM, Anthony Petrov wrote:
>>>>>
>>>>> Indeed, this is a nasty issue. Thanks for spotting it. You may notice
>>>>> that in most other places where something is logged, the code usually
>>>>> checks whether logging is enabled for a certain level of logging, e.g.:
>>>>>
>>>>>> if (shapeLog.isLoggable(PlatformLogger.FINER)) {
>>>>>> shapeLog.finer(
>>>>>> "*** INFO: Setting shape: PEER: " + this
>>>>>> + "; WINDOW: " + getWindow()
>>>>>> + "; TARGET: " + target
>>>>>> + "; SHAPE: " + shape);
>>>>>> }
>>>>>
>>>>> So basically, we could simply wrap the line in question into a similar
>>>>> if(){} statement to make sure it gets executed only when the FINEST
>>>>> level of logging is enabled. A similar solution may also apply to
>>>>> another call to enableLog.finer() in the same method just a few lines
>>>>> below.
>>>>
>>>> Although the fix has been already pushed to the workspace, let me
>>>> wonder if, in addition to wrapping the logging calls with the "if ()"
>>>> checks, we should have also rewritten isEventDisabled() to use
>>>> StringBuilder?
>>>
>>> I guess you really mean the XEvent.getFieldsAsString() here [1], right?
>>
>> Yes, exactly. Thanks for this correction.
>>
>>> Possibly. But with Federico's fix this method now only gets called when
>>> logging is enabled, and as such the impact of the issue is limited.
>>
>> I realize that. My point is that we just hide the problem, or make it less
>> visible, or make it appear less frequently - whatever you prefer. However,
>> the problematic code that contains char[] allocation is still in the
>> workspace, and we know how often people use the current code as a pattern
>> for future changes...
>>
>> Thanks,
>>
>> Artem
>>
>>> [1]
>>> http://mail.openjdk.java.net/pipermail/awt-dev/2011-October/001952.html
>>>
>>> --
>>> best regards,
>>> Anthony
>>>
>>>
>>>>
>>>> Thanks,
>>>>
>>>> Artem
>>>>
>>>>> Could you make a patch, test it, and post it to this mailing list for
>>>>> review please?
>>>>>
>>>>> --
>>>>> best regards,
>>>>> Anthony
>>>>>
>>>>> On 10/21/2011 7:10 AM, Federico Tello Gentile wrote:
>>>>>>
>>>>>> Hi.
>>>>>> I'm running
>>>>>>
>>>>>> java -version
>>>>>> java version "1.7.0_147-icedtea"
>>>>>> OpenJDK Runtime Environment (IcedTea7 2.0pre)
>>>>>> (7~b147-2.0~pre6-1ubuntu1)
>>>>>> OpenJDK 64-Bit Server VM (build 21.0-b17, mixed mode)
>>>>>>
>>>>>> and profiling a simple Swing application.
>>>>>>
>>>>>> Just by opening a JFrame and moving the mouse over it for 10 secods I
>>>>>> see 600 MB of char[] being created. I can easily create several
>>>>>> terabytes or those if I move the mouse a little longer. Thanks to the
>>>>>> incredibly efficient garbage collector the application performance is
>>>>>> not visibly affected on my 4GB quad core machine.
>>>>>>
>>>>>> The problem is an incredibly inefficient method
>>>>>> sun.awt.X11.XEvent.getFieldsAsString() which I pasted at the end of
>>>>>> this
>>>>>> message.
>>>>>>
>>>>>> The way it it handling string concatenation forces StringBuilder to
>>>>>> grow
>>>>>> many times and ends up calling Arrays.copyOf a lot of times.
>>>>>> It is being called by the sun.awt.X11.XWrapperBase.toString() just for
>>>>>> the sake of logging here:
>>>>>>
>>>>>> sun.awt.X11.XComponentPeer
>>>>>>
>>>>>> protected boolean isEventDisabled(XEvent e)
>>>>>>
>>>>>> enableLog.finest("Component is {1}, checking for disabled event
>>>>>> {0}", e,
>>>>>> (isEnabled()?"enabled":"disable"));
>>>>>>
>>>>>> The worse part is that even if logging is disabled and nothing at
>>>>>> all is
>>>>>> ever logged, the toString is called anyway and all those char[] are
>>>>>> created anyway.
>>>>>>
>>>>>> Here's a NetBeans profiler memory snapshot.
>>>>>>
>>>>>> http://ubuntuone.com/4xkprEzadUM4sUSdAnWlN5
>>>>>>
>>>>>>
>>>>>> This does not happen *at all* if I run the same profiling with this
>>>>>> openjdk version
>>>>>>
>>>>>> java -version
>>>>>> java version "1.6.0_23"
>>>>>> OpenJDK Runtime Environment (IcedTea6 1.11pre) (6b23~pre10-0ubuntu5)
>>>>>> OpenJDK 64-Bit Server VM (build 20.0-b11, mixed mode)
>>>>>>
>>>>>> I hope we can do something to improve this situation. I guess all Swing
>>>>>> applications are affected.
>>>>>>
>>>>>> I look forward for any comments.
>>>>>>
>>>>>> Here's the "thing"...
>>>>>>
>>>>>> ---------------------------------------------------
>>>>>>
>>>>>> String getFieldsAsString() {
>>>>>> String ret="";
>>>>>>
>>>>>> ret += ""+"type = " + XlibWrapper.eventToString[get_type()] +", ";
>>>>>> ret += ""+"xany = " + get_xany() +", ";
>>>>>> ret += ""+"xkey = " + get_xkey() +", ";
>>>>>> ret += ""+"xbutton = " + get_xbutton() +", ";
>>>>>> ret += ""+"xmotion = " + get_xmotion() +", ";
>>>>>> ret += ""+"xcrossing = " + get_xcrossing() +", ";
>>>>>> ret += ""+"xfocus = " + get_xfocus() +", ";
>>>>>> ret += ""+"xexpose = " + get_xexpose() +", ";
>>>>>> ret += ""+"xgraphicsexpose = " + get_xgraphicsexpose() +", ";
>>>>>> ret += ""+"xnoexpose = " + get_xnoexpose() +", ";
>>>>>> ret += ""+"xvisibility = " + get_xvisibility() +", ";
>>>>>> ret += ""+"xcreatewindow = " + get_xcreatewindow() +", ";
>>>>>> ret += ""+"xdestroywindow = " + get_xdestroywindow() +", ";
>>>>>> ret += ""+"xunmap = " + get_xunmap() +", ";
>>>>>> ret += ""+"xmap = " + get_xmap() +", ";
>>>>>> ret += ""+"xmaprequest = " + get_xmaprequest() +", ";
>>>>>> ret += ""+"xreparent = " + get_xreparent() +", ";
>>>>>> ret += ""+"xconfigure = " + get_xconfigure() +", ";
>>>>>> ret += ""+"xgravity = " + get_xgravity() +", ";
>>>>>> ret += ""+"xresizerequest = " + get_xresizerequest() +", ";
>>>>>> ret += ""+"xconfigurerequest = " + get_xconfigurerequest() +", ";
>>>>>> ret += ""+"xcirculate = " + get_xcirculate() +", ";
>>>>>> ret += ""+"xcirculaterequest = " + get_xcirculaterequest() +", ";
>>>>>> ret += ""+"xproperty = " + get_xproperty() +", ";
>>>>>> ret += ""+"xselectionclear = " + get_xselectionclear() +", ";
>>>>>> ret += ""+"xselectionrequest = " + get_xselectionrequest() +", ";
>>>>>> ret += ""+"xselection = " + get_xselection() +", ";
>>>>>> ret += ""+"xcolormap = " + get_xcolormap() +", ";
>>>>>> ret += ""+"xclient = " + get_xclient() +", ";
>>>>>> ret += ""+"xmapping = " + get_xmapping() +", ";
>>>>>> ret += ""+"xerror = " + get_xerror() +", ";
>>>>>> ret += ""+"xkeymap = " + get_xkeymap() +", ";
>>>>>> ret += "{" + get_pad(0) + " " + get_pad(1) + " " + get_pad(2) + " " +
>>>>>> get_pad(3) + " " + get_pad(4) + " " + get_pad(5) + " " + get_pad(6) + "
>>>>>> " + get_pad(7) + " " + get_pad(8) + " " + get_pad(9) + " " +
>>>>>> get_pad(10)
>>>>>> + " " + get_pad(11) + " " + get_pad(12) + " " + get_pad(13) + " " +
>>>>>> get_pad(14) + " " + get_pad(15) + " " + get_pad(16) + " " + get_pad(17)
>>>>>> + " " + get_pad(18) + " " + get_pad(19) + " " + get_pad(20) + " " +
>>>>>> get_pad(21) + " " + get_pad(22) + " " + get_pad(23) + " " + "}";
>>>>>> return ret;
>>>>>> }
>>>>>>
>>>>>>
>>
>



More information about the awt-dev mailing list