Some improvements needed on icedtea Java2D renderers.

[Xerxes Rånby]

Omair Majid skrev:
> Hi Xerxes,
>
> Xerxes Rånby wrote:
>> Hello, during bug testing of icedtea6 on various embedded and 
>> stationary hardware i have stumbled across some rendering issues of 
>> java2d when running under icedtea6.
>
> What version of icedtea6 are you using?
>
The pictures in my last mail with the pink issue boxes was made using 
the installed java on my Linux Mint machine based on ubuntu intrepid:
java version "1.6.0_0"
IcedTea6 1.3.1 (6b12-0ubuntu6.4) Runtime Environment (build 1.6.0_0-b12)
OpenJDK Server VM (build 1.6.0_0-b12, mixed mode)

Sorry for reporting bugs that had been fixed in the hg, i really thought 
i had generated the picture using at least 1.4.1.

I have attached a new picture how it all looks now when running Shark in 
interpreted mode using the icedtea hg tip, I had to make the pictures 
running with -Xint because i have not yet found the fix for pr324 that 
prevents the shark JIT to compile the whole testcase.
http://icedtea.classpath.org/bugzilla/show_bug.cgi?id=324

java version "1.6.0_0"
OpenJDK Runtime Environment (IcedTea6 1.5pre-r0cf247b95e6c) (build 
1.6.0_0-b16)
OpenJDK Shark VM (build 14.0-b10-product, mixed mode)

The only visible issues left that i can see now using this shark version 
are issue 1 and 2:

>> 1. When the window are refreshed, the component borders gets redrawn 
>> with a mix of soft and sharp borders.
This issue can be seen clearly at the borders of the Arcs_Curves tab on 
the attached picture.
>>
>> 2. The rendering of the black box have been requested to be rendererd 
>> using a dotted CAP_ROUND BasicStroke: it should be dotted!
>>        BasicStroke dotted = new BasicStroke(3, BasicStroke.CAP_ROUND,
>>                     BasicStroke.JOIN_ROUND, 0, new float[]{0,6,0,6}, 0);
>>        g2.setStroke(dotted);
>>        g2.drawRect(3,3,w-6,h-6);
> The other issue is the difference in how a length of 0 is treated by 
> the jdk. The comment in 
> icedtea6/openjdk/jdk/src/share/classes/sun/java2d/pisces/Dasher.java 
> line 34 points out the issue:
>  *
>  * <p> Issues: in J2Se, a zero length dash segment as drawn as a very
>  * short dash, whereas Pisces does not draw anything.  The PostScript
>  * semantics are unclear.
>  *
> So the above code in JSE would draw a dashed border but icedtea6 would 
> draw a blank border. "new float[]{0.01,6,0.01,6}" should work.
>
I agree it  are unclear how this should be fixed. Thank you for insight 
into the complexity of the zero sized dach.


Issue 3 with the reversed spaces/dach as you had pointed out had already 
been fixed in the mercurial hg, thank you.

>> 4. The panel are not resized properly to fit the slider component.
> Looks like the components above are taking up too much vertical space. 
> This looks like bug 255 [3]. I have not been able to reproduce this 
> locally. Any help reproducing this is appreciated.
Fixed for me as well using the mercurial icedtea6 build!
looks ok, perhaps the panel are now slightly larger than the container, 
it hides the bottom border yet this could be a examplecode error.

>>
>> 5. and 6.  The last two issues are related. the specified joinType 
>> are not applyed to lines that gets inserted using closePath.
>>    public void drawDemo(int w, int h, Graphics2D g2) {
>>        BasicStroke bs = new BasicStroke(bswidth,
>>                                    BasicStroke.CAP_BUTT, joinType);
>>        GeneralPath p = new GeneralPath();
>>        p.moveTo(- w / 4.0f, - h / 12.0f);
>>        p.lineTo(+ w / 4.0f, - h / 12.0f);
>>        p.lineTo(- w / 6.0f, + h / 4.0f);
>>        p.lineTo(+     0.0f, - h / 4.0f);
>>        p.lineTo(+ w / 6.0f, + h / 4.0f);
>>        p.closePath();                                          
>> //      <-----  issue 6: this should have create a correct join. 
>> instead a round join are allways used.
>>        p.closePath();                                          
>> //      <-----  issue 5: this should have create a correct join. the 
>> rendering of the start line should
>>                                                                        
>> //                have  been updated when the second closepath was 
>> instructed.
Like you i cant really understand the second p.closePath() in the 
example code, the more I look at it and the the more I compare sun jdk 
and icedtea only indicate that it are not doing anything at all!

Your recent closePath fixes for issue 5 and 6 worked like a charm for 
me, thank you Omair it simply looks great!

>>        g2.translate(w/2, h/2);
>>        g2.setColor(Color.black);
>>        g2.draw(bs.createStrokedShape(p));
>>    }
>>
>>
>
> Cheers,
> Omair
> [3] http://icedtea.classpath.org/bugzilla/show_bug.cgi?id=255

Cheers and have a great day!
Xerxes
-------------- next part --------------
A non-text attachment was scrubbed...
Name: sharkXintIssue1and2Left-Java2Demo.png
Type: image/png
Size: 61139 bytes
Desc: not available
Url : http://mail.openjdk.java.net/pipermail/distro-pkg-dev/attachments/20090430/86058adf/sharkXintIssue1and2Left-Java2Demo.png