RFR: 8370141: [macOS] Crash after PrinterJob ends when Graphics.create() is used. [v2]

[Phil Race]

On Fri, 24 Oct 2025 18:30:13 GMT, Phil Race <prr at openjdk.org> wrote:

>> multi-threaded version of the `PrintJobAfterEndTest` always crashed for me even with a patch:
>> 
>> 
>> import java.awt.Frame;
>> import java.awt.Graphics;
>> import java.awt.JobAttributes;
>> import java.awt.JobAttributes.DialogType;
>> import java.awt.PageAttributes;
>> import java.awt.PrintJob;
>> import java.awt.Toolkit;
>> import java.util.concurrent.CountDownLatch;
>> 
>> public final class MTPrintJobAfterEndTest {
>> 
>>     public static void main(String[] args) throws InterruptedException {
>> 
>>         JobAttributes jobAttributes = new JobAttributes();
>>         jobAttributes.setDialog(DialogType.NONE);
>>         PageAttributes pageAttributes = new PageAttributes();
>>         Frame f = new Frame();
>>         Toolkit toolkit = f.getToolkit();
>> 
>>         for (int i = 0; i < 1000; i++) {
>>             PrintJob job = toolkit.getPrintJob(f, "Crash Test",jobAttributes,
>>                                                pageAttributes);
>>             if (job != null) {
>>                 Graphics g = job.getGraphics();
>>                 CountDownLatch latch = new CountDownLatch(1);
>>                 
>>                 Thread endThread = new Thread(() -> {
>>                     try {
>>                         latch.await();
>>                         job.end();
>>                     } catch (Throwable ignore) {}
>>                 });
>>                 
>>                 Thread drawThread = new Thread(() -> {
>>                     try {
>>                         latch.await();
>>                         g.drawLine(0, 100, 200, 100);
>>                     } catch (Throwable ignore) {}
>>                 });
>>                 
>>                 endThread.start();
>>                 drawThread.start();
>>                 latch.countDown();
>>                 
>>                 endThread.join();
>>                 drawThread.join();
>>             }
>>         }
>>     }
>> }
>
> I have updated the fix making the SurfaceData methods synchronized and wrapped the disposed with synchronizing on the same surface.
> So now it is not possible to have the state changed whilst executing any of these methods.
> This fixes the MT crash. I've updated both the tests to use MT and also added more primitives being drawn.

Meanwhile, I tried the MT test on Windows and I got a crash there too.
It took me a while but I think a similar approach is needed there and seems to work.
I'm going to create a separate bug ID for that as it is entirely windows-specific changes and this is (test aside) entirely Mac-specific changes.

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

PR Review Comment: https://git.openjdk.org/jdk/pull/27905#discussion_r2461609072