[GreenFuze]

Fixes #3665.

As you suggested, initFrame no longer takes className. None of the natives ever read it (each one always creates the platform's default embedded frame class), and the GTK native didn't even declare it.

SWT_AWT.embeddedFrameClass is now deprecated for removal (since 2026-12), following the API removal process. I kept its Class.forName check, so nothing changes for code that still sets it. On Cocoa, JDK17_FRAME was only used as that default, so it's gone too.

Like in #3526, I didn't touch Library.java or the binaries -- the C change needs the natives rebuilt. Happy to adjust anything!

[github-actions]

Test Results

 1 026 files  ±0   1 026 suites  ±0   1h 34m 15s ⏱️ + 9m 48s
 5 067 tests ±0   5 039 ✅ ±0     28 💤 ±0  0 ❌ ±0 
29 761 runs  ±0  27 225 ✅ ±0  2 536 💤 ±0  0 ❌ ±0 

Results for commit 779e7a5. ± Comparison against base commit fcc15be.

♻️ This comment has been updated with latest results.

[GreenFuze]

Heads up: API tools treats adding @noreference to an existing public field as an API break (it reports the field as "no longer an API" and wants a major version bump), which is why the builds failed. So I dropped @noreference and kept the @Deprecated(forRemoval = true, since = "2026-12") deprecation. If you'd rather keep @noreference, I can add API filters to the binary fragments instead, just say.

[GreenFuze]

All GitHub builds pass now. The Jenkins job timed out twice at ~2h, I think in the native-build stage (this PR touches the cocoa and win32 natives). Could someone re-run it when the native build agents are free?

[akurtakov]

Please change the natives too. The C part will be auto rebuilt by the CI.

[GreenFuze]

@akurtakov both points are in:

• @noreference is back on embeddedFrameClass, with an API tools filter (id 338940029) in each binary fragment's .api_filters (5165be9). My inline reply said cc2c5a7; that commit became 5165be9 when I rebased.
• The natives are changed: the cocoa and win32 swt_awt.c no longer take className (813566f), and the GTK native never declared it. I left Library.java and the binaries for the CI rebuild, as you said.

All GitHub builds are green. The Jenkins run on the current head was aborted again after about 2 hours, like the earlier ones. Could someone re-run it, or let me know if there's anything else on my side?

[akurtakov]

Rebasing on top of master does all you have asked for, I did it now. I would appreciate if you squash all the comments into single commit.

[akurtakov]

Sorry for the delay! There are so many things ongoing and so less times that PRs easily get lost.

[akurtakov]

So far this is blocked by infrastructure problem https://gitlab.eclipse.org/eclipsefdn/helpdesk/-/work_items/8008 . If natives on windows can't be rebuild I can't merge this change