Graphics2D.drawString() fails with Font derived from rotated + scaled AffineTransform

[Daniel Gredler]

Hi all,

A bug number was issued for this issue [1], and I've created the
corresponding PR [2]. Can a couple of reviewers spend a few cycles
reviewing it? The fix is relatively simple, I think the main questions are
more stylistic / cultural (simplicity vs. performance, compatibility
constraints).

Thanks!

Daniel

[1] https://bugs.openjdk.org/browse/JDK-8339974
[2] https://github.com/openjdk/jdk/pull/20993


On Tue, Sep 10, 2024 at 1:56 AM Daniel Gredler <djgredler at gmail.com> wrote:

> Hi all,
>
> I've done a little more digging, and there is indeed a deeper bug in
> TextLayout.getOutline(AffineTransform), where the transformation is applied
> too early (before the layout path applies its mapping).
>
> There are two uses of TextLayout.getOutline(AffineTransform): the first
> one, triggered via a simple Graphics2D.drawString( ) was the one I ran
> into; the second one requires the use of a transformed font with a
> PSPrinterJob. Both are broken, so it doesn't really make sense to work
> around the TextLayout issue for the first scenario, but leave the second
> scenario broken.
>
> The deeper fix, including two test classes verifying that both scenarios
> were broken and now work correctly, is available on GitHub [1]. I plan to
> submit a PR as soon as I have a final bug ID from my web bug submission.
>
> Two points I'd like some feedback on:
> 1. I've changed the signature of method
> TextLine.getOutline(AffineTransform), to just TextLine.getOutline( ); I'm
> assuming this is OK since TextLine is a package-private class?
> 2. I've added a fast path in TextLayout.getOutline(AffineTransform) to
> transform the shape in place if it is a GeneralPath (it should usually be,
> and avoids an extra Shape copy vs. the current code); does this look OK? Or
> should the code be kept completely generic and a copy always made? I had a
> look at just changing the variable type from Shape to GeneralPath, but the
> change started to cascade into a few other classes and I pulled the plug.
>
> Take care,
>
> Daniel
>
> [1]
> https://github.com/openjdk/jdk/compare/master...gredler:jdk:draw-string-bug-2
>
>
>
> On Thu, Sep 5, 2024 at 8:30 PM Daniel Gredler <djgredler at gmail.com> wrote:
>
>> Hi all,
>>
>> I've run into a bug [1] in the drawing of strings via
>> Graphics2D.drawString(), when used with a Font derived from a rotated +
>> scaled AffineTransform. If the AffineTransform scaling is small, the text
>> draws correctly. But if the scale is too large, the text does not draw. [3]
>>
>> It looks like it's an effect of the OutlineTextRenderer.THRESHHOLD; text
>> larger than this threshold is sent to the OutlineTextRenderer and fails to
>> draw. The fix [2] seems to be to apply the position translation
>> transformation *after* the text outline shape is created, rather than
>> passing it into TextLayout.getOutline(). However, I do not know if this is
>> the correct fix, or if it is just a workaround to a deeper issue in
>> TextLayout.getOutline(), which in theory might need to wait to apply the
>> provided translation until after the layout path has had a chance to map to
>> user space.
>>
>> Can someone familiar with this code comment on whether the proposed fix
>> is the correct fix, or whether deeper surgery is needed in
>> TextLayout.getOutline()?
>>
>> My other small worry with the current fix [2] is that it creates an extra
>> copy of the Shape object, so it's going to allocate a bit more memory than
>> the current implementation. Not sure if there's a better way to update the
>> Shape in-place.
>>
>> Take care,
>>
>> Daniel
>>
>> ---
>>
>> [1] Oracle internal review ID 9077538, no public bug ID yet
>> [2]
>> https://github.com/openjdk/jdk/compare/master...gredler:jdk:draw-string-bug
>> [3]
>> import java.awt.Color;
>> import java.awt.Font;
>> import java.awt.Graphics2D;
>> import java.awt.geom.AffineTransform;
>> import java.awt.image.BufferedImage;
>> import java.io.File;
>>
>> import javax.imageio.ImageIO;
>>
>> /**
>>  * <p>The "TEST 2" text does not draw. NOTE: Reducing the AffineTransform
>> scale
>>  * from 12 to 10 seems to allow it to draw?!?
>>  *
>>  * <p>Oracle internal review ID : 9077538
>>  */
>> public class DrawStringTest {
>>
>>     public static void main(String[] args) throws Exception {
>>
>>         AffineTransform at = AffineTransform.getRotateInstance(3 *
>> Math.PI / 2);
>>         at.scale(12, 12);
>>
>>         Font font1 = new Font("SansSerif", Font.PLAIN, 10);
>>         Font font2 = font1.deriveFont(at);
>>
>>         BufferedImage image = new BufferedImage(500, 500,
>> BufferedImage.TYPE_BYTE_GRAY);
>>         Graphics2D g2d = image.createGraphics();
>>         g2d.setColor(Color.WHITE);
>>         g2d.fillRect(0, 0, image.getWidth(), image.getHeight());
>>         g2d.setColor(Color.BLACK);
>>         g2d.setFont(font1);
>>         g2d.drawString("TEST 1", 200, 400); // draws text
>>         g2d.setFont(font2);
>>         g2d.drawString("TEST 2", 200, 400); // does not draw text
>>         g2d.dispose();
>>
>>         ImageIO.write(image, "png", new File("test.png"));
>>     }
>>
>> }
>>
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.openjdk.org/pipermail/client-libs-dev/attachments/20240917/f1fc15aa/attachment.htm>