No problem! Thanks for rechecking the issue as well, and that you did some investigations beforehand. Everything helps!
In the meantime, I also forked the project and checked the original ControlsFX code.
Just when I verified my assumptions and fix, you created the PR! Thanks, I already reviewed it.
-- Marius
Gesendet: Montag, 24. November 2025 um 20:37
Von: "Cormac Redmond" <[email protected]>
An: "Daniel Peintner" <[email protected]>
CC: "Marius Hanl" <[email protected]>,[email protected]
Betreff: Re: JFX 26-ea+15 TableView "bug" -- related to 8359599?
Hi both,
Thank you Marius for investigating this, and apologies for assuming it was a JFX bug (I did spend time trying to determine if it was ControlsFX misbehaving).
Thank you Daniel for following up. I created a PR on ControlsFX: https://github.com/controlsfx/controlsfx/pull/1607 ...
Kind Regards,
Cormac
On Mon, 24 Nov 2025 at 15:50, Daniel Peintner <[email protected]> wrote:
Hi,Thank you both for reporting the problem and providing a solution.I opened https://github.com/controlsfx/controlsfx/issues/1606 to keep track of the problem.Marius, I think it would be good if you can provide a fix/PR.Note: I am not one of the maintainers of ControlsFX but a user and I think it is good if we can resolve the problem.Thanks,-- Daniel
On Mon, Nov 24, 2025 at 6:50 AM Marius Hanl <[email protected]> wrote:The refresh changes could only be a problem if something in the Code is misbehaving.I had a look, and indeed, the expandedNode is NEVER removed. The logic simply does not work, never did in the previous versions.Prior to 26-ea+15, the table refresh was recreating ALL cells. This masked the bug.But now, the refresh method really refreshes the existing cells. So rows and cells need to react to that, as they should before.The root cause is, that the expandedNode is removed from the HashMap before it could be cleaned up in the corresponding TableRowSkin.So when the TableRowSkin tries to remove the expandedNode, it is already null.Removing the code fixes the issue:public BooleanProperty getExpandedProperty(S item) {
BooleanProperty value = expansionState.get(item);
if (value == null) {
value = new SimpleBooleanProperty(item, "expanded", false) {@Override
protected void invalidated() {
// if (!getValue()) {
// expandedNodeCache.remove(getBean());
// }
getTableView().refresh();
}
};
expansionState.put(item, value);
}
return value;
}Again, this was also problematic before, but masked by the recreation.To restore the old behavior, you could write:TableView<Person> table = new TableView<>() {
@Override
public void refresh() {
getProperties().put("recreateKey", Boolean.TRUE);
}
};I don't know if ControlsFX is still somewhat maintained. Otherwise I can have file a PR there.-- MariusGesendet: Sonntag, 23. November 2025 um 16:14Von: "Cormac Redmond" <[email protected]>Betreff: JFX 26-ea+15 TableView "bug" -- related to 8359599?Hi,Since 26-ea+15, there may be a "bug" with TableView which appears to be rendering duplicated nodes.I first noticed this when using ControlFX's TableRowExpanderColumn (https://controlsfx.github.io/javadoc/11.2.2/org.controlsfx.controls/org/controlsfx/control/table/TableRowExpanderColumn.html), which has been working as expected for several years, right up until 26-ea+15. This column just provides a simple way to allow expansion of table rows.I have created a gist consisting of a single runnable class. It's quite small, and no dependencies are required (minimal ControlsFX code copied into it):Run this on 26-ea+15, and you'll see the issue re: duplicated rendering; run it on anything older, and there's no issue. If I had to guess, I'd say this issue is related to this commit (8359599):The application starts out like this on all versions, there's three rows (number of rows is irrelevant):Click the +'s and -'s more than once.26-ea+15 "bug"; notice the duplications that appear just by clicking the + / -'s a couple of times:26-ea+14 (and previous) expected / normal behaviour:Kind Regards,Cormac Redmond
