Trying to clear a feature typing using the reference widget does not work

[pcdavid]

Steps to reproduce

1. Create a SysML project and a GV diagram
2. Create a Part, with a typed attribute, eg. "volume : Real"
3. Make sure the attribute is selected
4. In the details view, the "Typed by" section at the bottom shows the attribute's type and proposes two visible ways to clear is: through the "x" on the chip, and through the "Clear" button.
5. Try both.

Expected result

The attribute's type is cleared, the model is in the same state as when "direct editing" the attribute with ":".

Actual result

Both attempts fail after a few seconds with a timeout error.

On the backend:

Caused by: java.lang.UnsupportedOperationException
        at org.eclipse.emf.common.util.BasicEList$UnmodifiableEList.remove(BasicEList.java:947) ~[org.eclipse.emf.common-2.31.0.jar!/:na]
        at org.eclipse.sirius.components.view.emf.widget.reference.ReferenceWidgetBehaviorConverter.handleItemRemove(ReferenceWidgetBehaviorConverter.java:112) ~[sirius-components-view-emf-widget-reference-2026.1.0.jar!/:2026.1.0]
        at org.eclipse.sirius.components.view.emf.widget.reference.ReferenceWidgetBehaviorConverter.lambda$convert$1(ReferenceWidgetBehaviorConverter.java:64) ~[sirius-components-view-emf-widget-reference-2026.1.0.jar!/:2026.1.0]
        at org.eclipse.sirius.components.widget.reference.ReferenceWidgetComponent.lambda$getItems$6(ReferenceWidgetComponent.java:149) ~[sirius-components-widget-reference-2026.1.0.jar!/:2026.1.0]
        at java.base/java.util.Optional.map(Optional.java:260) ~[na:na]
        at org.eclipse.sirius.components.collaborative.widget.reference.handlers.RemoveReferenceValueEventHandler.handle(RemoveReferenceValueEventHandler.java:96) ~[sirius-components-collaborative-widget-reference-2026.1.0.jar!/:2026.1.0]

The stack mentions 2026.1.0 where I reproduced it to make sure it was not a regression, but this is also present in 2026.3.8.

[pcdavid]

Capture.video.du.2026-05-07.10-09-23.mp4

[pcdavid]

This used to work in 2025.10.0, but fails in 2025.12.0 and later.

[pcdavid]

Maybe related to 386289c?

[AxelRICHARD]

Maybe related to 386289c?

yes it is