RFR: 8347937: Canvas pattern test fails on WebKit 620.1

Sergey Nazarkin snazarki at openjdk.org
Fri Mar 14 17:21:00 UTC 2025


On Fri, 14 Mar 2025 14:12:30 GMT, Kevin Rushforth <kcr at openjdk.org> wrote:

>> Hi!
>> 
>> This is my desperate fix for what I believe to be a critical bug. Since my knowledge of the codebase is minimal, and the class structure is a bit controversial, my fix is mostly based on historical excavation.
>> 
>> #  What is the problem?
>> VM started to crash after webkit 620.1 update when running simple [CanvasTest](https://github.com/openjdk/jfx/blob/1c3cfcb8bb4e8874406d3a5b507f38859f4c1d9b/modules/javafx.web/src/test/java/test/javafx/scene/web/CanvasTest.java#L125).  Running with `-Xcheck:jni ` reveals invalid object is passed to JNI [call](https://github.com/openjdk/jfx/blob/0555fb25a16b6b6705a42c6d8592cf1c6ddccc67/modules/javafx.web/src/main/native/Source/WebCore/platform/graphics/java/NativeImageJava.cpp#L56): expected is`WCImageFrame` but provided is `RTImage` (an descendant of `WCImage`). The latter has no `getSize` method.
>> Stack trace
>> 
>>     frame #4: 0x0000000101a36094 libjvm.dylib`checked_jni_CallObjectMethodV(env=0x0000000124a89cb0, obj=0x000000012310f9e8, methodID=0x00006000019abde8, args="\xa0\U00000012\U0000001ev\U00000001") at jniCheck.cpp:1012:1 [opt]
>>     frame #5: 0x000000033e88ea60 libjfxwebkit.dylib`JNIEnv_::CallObjectMethod(this=0x0000000124a89cb0, obj=0x000000012310f9e8, methodID=0x00006000019abde8) at jni.h:906:18
>>     frame #6: 0x00000003426fcda0 libjfxwebkit.dylib`WebCore::PlatformImageNativeImageBackend::size(this=0x0000600000e12790) const at NativeImageJava.cpp:55:48
>>     frame #7: 0x000000034258d76c libjfxwebkit.dylib`WebCore::NativeImage::size(this=0x00000001761c2c00) const at NativeImage.cpp:89:23
>>     frame #8: 0x00000003425723c0 libjfxwebkit.dylib`WebCore::ImageFrame::ImageFrame(this=0x00000001761e1270, nativeImage=0x000000016fdf7090) at ImageFrame.cpp:40:29
>>     frame #9: 0x0000000342572420 libjfxwebkit.dylib`WebCore::ImageFrame::ImageFrame(this=0x00000001761e1270, nativeImage=0x000000016fdf7090) at ImageFrame.cpp:39:1
>>     frame #10: 0x00000003425a4d7c libjfxwebkit.dylib`WebCore::NativeImageSource::NativeImageSource(this=0x00000001761e1260, nativeImage=0x000000016fdf7090) at NativeImageSource.cpp:37:7
>>     frame #11: 0x00000003425a4d24 libjfxwebkit.dylib`WebCore::NativeImageSource::NativeImageSource(this=0x00000001761e1260, nativeImage=0x000000016fdf7090) at NativeImageSource.cpp:38:1
>>     frame #12: 0x00000003425a4cac libjfxwebkit.dylib`WebCore::NativeImageSource::create(nativeImage=0x000000016fdf7090) at NativeImageSource.cpp:33:26
>>     frame #13: 0x00000003423a6f18 libjfxwebkit.dylib`WebCore::B...
>
> Initial testing shows that while this does prevent the crash, it doesn't actually render anything in the failing case.
> 
> @jaybhaskar will add more detail on Monday. He and @Gopalora are working on an actual fix.

@kevinrushforth thank you for the review. I expected something like this. BTW, we've just found that win32 and osx-x64 crashes even with this fix.

-------------

PR Comment: https://git.openjdk.org/jfx/pull/1734#issuecomment-2725303486


More information about the openjfx-dev mailing list