Add workflow to run with GTK4 enabled

[laeubi]

Currently everything is run with GTK3 by default but we can't tell the current state of affairs with GTK4.

This now adds a new action that run the build (and tests) with GTK4 enabled.

[github-actions]

Test Results

   502 files  ±0     502 suites  ±0   9m 36s ⏱️ +33s
 4 334 tests ±0   4 320 ✅ ±0   14 💤 ±0  0 ❌ ±0 
16 575 runs  ±0  16 466 ✅ ±0  109 💤 ±0  0 ❌ ±0 

Results for commit 7e158d3. ± Comparison against base commit 857ac0b.

♻️ This comment has been updated with latest results.

[laeubi]

If one goes the summary tab here:

https://github.com/eclipse-platform/eclipse.platform.swt/actions/runs/13304656046?pr=1663

then one can download the test-results that now also contain an crash report
hs_err_pid5270.log

First lines

Current thread (0x00007f0eec029c90):  JavaThread "main" [_thread_in_native, id=5271, stack(0x00007f0ef0b00000,0x00007f0ef0c00000)]

Stack: [0x00007f0ef0b00000,0x00007f0ef0c00000],  sp=0x00007f0ef0bfbe70,  free space=1007k
Native frames: (J=compiled Java code, j=interpreted, Vv=VM code, C=native code)
C  [libgobject-2.0.so.0+0x26691]
C  [libgobject-2.0.so.0+0x27fc4]  g_object_new_with_properties+0x1d4
C  [libgobject-2.0.so.0+0x28f71]  g_object_new+0xc1
C  [libgtk-4.so.1+0x237c45]
C  [libgobject-2.0.so.0+0x40ff9]  g_type_create_instance+0x359
C  [libgobject-2.0.so.0+0x26a64]
C  [libgobject-2.0.so.0+0x28016]  g_object_new_with_properties+0x226
C  [libgobject-2.0.so.0+0x28f71]  g_object_new+0xc1
C  [libgtk-4.so.1+0x12ca33]
C  [libgobject-2.0.so.0+0x40ff9]  g_type_create_instance+0x359
C  [libgobject-2.0.so.0+0x26a64]
C  [libgobject-2.0.so.0+0x28016]  g_object_new_with_properties+0x226
C  [libgobject-2.0.so.0+0x28f71]  g_object_new+0xc1
C  [libswt-pi4-gtk-4968r11.so+0x3be1c]  Java_org_eclipse_swt_internal_gtk_GTK_gtk_1entry_1new+0x9
j  org.eclipse.swt.widgets.Display.initializeSystemColorsDisabled()V+18
j  org.eclipse.swt.widgets.Display.initializeSystemColors()V+59
j  org.eclipse.swt.widgets.Display.init()V+13
j  org.eclipse.swt.graphics.Device.<init>(Lorg/eclipse/swt/graphics/DeviceData;)V+145
j  org.eclipse.swt.widgets.Display.<init>(Lorg/eclipse/swt/graphics/DeviceData;)V+2
j  org.eclipse.swt.widgets.Display.<init>()V+2
j  org.eclipse.swt.tests.junit.Test_org_eclipse_swt_widgets_Display.test_syncCall_dispose()V+4
v  ~StubRoutines::call_stub

[ptziegler]

That looks very promising! 😄

I've created #1832 to solve the current segmentation fault, but I fear there are a lot of other crashes just waiting to be discovered.

This obviously can't be merged yet, otherwise there won't be a green build for the next few years. But I still like the idea of getting feedback whether my (or others) GTK4 changes cause non-trivial problems. So what do you think the next steps should be?

[laeubi]

This obviously can't be merged yet, otherwise there won't be a green build for the next few years.

If you find this useful I can further work on this, one idea could be that we run this job if you add a label to a PR (e.g. gtk4) then it can already be used but won't influence "regular" PRs.

[laeubi]

Closing in favor of:

• Run the Github verification with GTK4 enabled if it has GTK4 label #1833
  you can now add the gtk4 label to a PR to run the build/test of SWT with GTK4.

[HannesWell]

* [Run the Github verification with GTK4 enabled if it has GTK4 label #1833](https://github.com/eclipse-platform/eclipse.platform.swt/pull/1833)
  you can now add the `gtk4` label to a PR to run the build/test of SWT with GTK4.

Maybe a naive question, but can't we have another branch in the maven-build matrix for Linux-gtk4?

[laeubi]

Maybe a naive question, but can't we have another branch in the maven-build matrix for Linux-gtk4?

Basically yes, but as GTK4 currently always fails this would result in each an every PR having failed checks (what is not really desirable), so given GTK4 works at some point in time and we still want to retain GTK3 that would be an option... on the other hand SWT has always ever only supported one main line as far as I know.