Possible regression between IcedTea6-1.8 and IcedTea6-1.9

[Pavel Tisnovsky]

Pavel Tisnovsky wrote:
> Dr Andrew John Hughes wrote:
>> On 18:27 Thu 12 Aug     , Pavel Tisnovsky wrote:
>>> Hi all,
>>>
>>> I tried to find out why some new regression tests does not pass on the
>>> IcedTea6 HEAD and - by the way - I may have found the regression between
>>> IcedTea6 IcedTea6-1.8 and-1.9.
>>>
>>> When the following very simple test is run with option
>>> -Dsun.java2d.opengl set to True and I run it inside VNC session (ie
>>> without acceleration):
>>>
>>> import javax.swing.JPanel;
>>>
>>> public class Test extends JPanel {
>>>     public Test() {
>>>     }
>>>
>>>     public static void main(String[] args)
>>>     {
>>>         Test t = new Test();
>>>         System.out.println("Test Passed.");
>>>     }
>>> }
>>>
>>> It simply wrote warning message in IcedTea6-1.8:
>>>
>>> /temp/icedtea6-1.8/openjdk/build/linux-amd64/j2sdk-image/bin/java
>>> -Dsun.java2d.opengl=True Test
>>> Could not enable OpenGL pipeline for default config on screen 0
>>> Test Passed.
>>>
>>> but it failed in IcedTea6-1.9 (and HEAD too):
>>>
>>> /temp/icedtea6-1.9/openjdk.build/j2sdk-image/bin/java
>>> -Dsun.java2d.opengl=True Test
>>> Exception in thread "main" java.lang.UnsatisfiedLinkError:
>>> sun.java2d.opengl.OGLContext.getOGLIdString()Ljava/lang/String;
>>>         at sun.java2d.opengl.OGLContext.getOGLIdString(Native Method)
>>>         at
>>> sun.java2d.opengl.GLXGraphicsConfig$1.run(GLXGraphicsConfig.java:126)
>>>         at
>>> sun.java2d.opengl.OGLRenderQueue$QueueFlusher.run(OGLRenderQueue.java:220)
>>> ~
>>>
>>>
>>> Tested on RHEL5 x86_64 in VNC session, DISPLAY is set to :2.0 (it's
>>> correct, of course).
>>>
>>> Do anyone remember that we did any changes in XRender or java2d.opengl
>>> packages between 1.8 and 1.9?
>>>
>>> PS: at least these regression tests fail due to issue mentioned above:
>>> FAILED: sun/java2d/DirectX/AcceleratedScaleTest/AcceleratedScaleTest.java
>>> FAILED: sun/java2d/DirectX/AccelPaintsTest/AccelPaintsTest.java
>>> FAILED:
>>> sun/java2d/DirectX/OnScreenRenderingResizeTest/OnScreenRenderingResizeTest.java
>>> FAILED: sun/java2d/DirectX/OverriddenInsetsTest/OverriddenInsetsTest.java
>>> FAILED:
>>> sun/java2d/DirectX/StrikeDisposalCrashTest/StrikeDisposalCrashTest.java
>>> FAILED: sun/java2d/DirectX/TransformedPaintTest/TransformedPaintTest.java
>>> FAILED: sun/java2d/OpenGL/DrawBufImgOp.java
>>> FAILED: sun/java2d/OpenGL/GradientPaints.java
>>> FAILED: sun/java2d/pipe/hw/RSLAPITest/RSLAPITest.java
>>> FAILED: sun/java2d/pipe/MutableColorTest/MutableColorTest.java
>>>
>>> Cheers
>>> Pavel
>> As discussed on the mailing list, XRender was updated to the new version:
>>
>> http://icedtea.classpath.org/hg/release/icedtea6-1.9/rev/9da4f50c13c0
>>
>> Part of this involved updating the pipeline code which XRender relies on
>> (see the patch).  Both are 7 backports.  Have you checked whether the
>> same tests fail in 7?
> 
> I think that this issue could be solved by the patch stored in
> attachment [OGLContext.patch] because OGLContext.o is already included
> in xawt.so (which is correct) but it does not exist JNI link to native
> function OGLContext.getOGLIdString().
> 
> Is it possible to add this patch to IcedTea6 or (preferably) add its
> contents to existing patch 6961633-xrender-02.patch?
> 
> But I found another issue: when I run the test against IcedTea6 with the
> patch applied JRE failed due to fatal error - see attachment
> [hs_err_pid32222.log]. It's quite strange for me because libGL.so
> contains function glGetString(). I have to further investigate why this
> call failed (I run the test in VNC session which can be one possible
> source of failure).

So after many hours of testing:

This failure happens on RHEL 5 box, but not (for example) on Fedora 10.
I tried to run this very simple test in VNC sessions on both OSs:

#include <GL/gl.h>
#include <stdlib.h>
#include <stdio.h>

int main(void)
{
    char *c;
    glGetString(GL_VENDOR);
    c=(char*)glGetString(GL_VERSION);
    printf("%x\n", (int)c);
    return 0;
}

In VNC session which was started in Fedora 10 this test prints only 0
(ie empty string) which is fine because OpenGL is not initialized and
possibly can't be initialized. But the same test SEGFAULTed on RHEL 5
(also in VNC session).

Maybe we could add some simple test if OpenGL can be initialized, for
example something like this:
Display                 *dpy;
Window                  root;
XVisualInfo             *vi;

 dpy = XOpenDisplay(NULL);
 root = DefaultRootWindow(dpy);
 vi = glXChooseVisual(dpy, 0, att);
 if (!vi) {print_error_message_or_disable_opengl}