Hello,

I analyzed a little more our problem. It concerns only one SVG file. And that is a filter that causes problem:
<filter
        inkscape: collect = "always"
        id = "filter3834">
       <feGaussianBlur
          inkscape: collect = "always"
          stdDeviation = "2.2770497"
          id = "feGaussianBlur3836" />
</ filter>
If I replace it by
<filter
        inkscape: collect = "always"
        id = "filter3834">
</ filter>
there is no more problem. I do not know if this information is sufficient to reopen the issue https://bugs.openjdk.java.net/browse/JDK-8019201 on JDK side. But anyway, I do not have the rights to do so.

I join the complete SVG file for information and the error stack.

Regards,

Laurent

Le 25/09/2013 19:11, Laurent Redor a écrit :
Hello,

The bug https://bugs.openjdk.java.net/browse/JDK-8019201 is fixed. And effectively, if I run the test includes in the commit http://hg.openjdk.java.net/jdk8/2d/jdk/rev/833f05116f7b :

However, I have always similar problem with JDK 1.7.0_40-b43 and the AffineTransformOp JDK class.
I do not know enough about the Batik library and even less ConvolveOp and AffineTransformOp JDK classes to adapt the test to reveal the problem with AffineTransformOp.

Does someone who knows enough these classes could adapt the test?


Best regards,

Laurent

>> Hi everyone,
>> 
>> the latest security update of Oracle's JDK/JRE 1.7.0_25 breaks many if 
>> not all BufferedImageOp.filter implementations.
>> 
>> BATIK 1.7 uses these operations in several places
>>    sources/org/apache/batik/ext/awt/g2d/AbstractGraphics2D.java
>>  
>> sources/org/apache/batik/ext/awt/image/renderable/ConvolveMatrixRable8Bit.java
>>  
>> sources/org/apache/batik/ext/awt/image/renderable/MorphologyRable8Bit.java
>>    sources/org/apache/batik/ext/awt/image/rendered/AffineRed.java
>>    sources/org/apache/batik/ext/awt/image/rendered/Any2LsRGBRed.java
>>    sources/org/apache/batik/ext/awt/image/rendered/Any2LumRed.java
>>    sources/org/apache/batik/ext/awt/image/rendered/Any2sRGBRed.java
>>    sources/org/apache/batik/ext/awt/image/rendered/ProfileRed.java
>>    sources/org/apache/batik/svggen/SVGGraphics2D.java
>> and consequently BATIK may break down with unexpected 
>> "java.awt.image.ImagingOpException: Unable to transform src image" 
>> exceptions.
>> 
>> The only workaround I have been able to find so far is patching BATIK 
>> sources to no longer call BufferedImageOp.filter directly on the source 
>> and destination image, but to filter two ARGB BufferedImage copies.
>> 
>> A corresponding bug report 
>> http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=8019201 has been 
>> filed with Oracle. However, the report does not mention the full extend 
>> of this problem - since filing the report, I have found Linux JDKs and 
>> additional BufferedImageOp implementations such as AffineTransformOp 
>> suffering from this problem, too.
>
>A project of mine has had two reports of this very probelm from users
>using 1.7.0_25 on Windows, and I can reliably reproduce it myself with
>OpenJDK 1.7.0_25 on Linux.
>
>Is there any way we can poke Oracle to fix the regression faster?
>
>-- 
>J.