On Fri, 16 Oct 2026 18:37:11 GMT, Andy Goryachev <[email protected]> wrote:
>> This PR fixes a regression from https://github.com/openjdk/jfx/pull/1830. >> >> My suspicion was that this might be a preexisting after all, which just >> happens to be much more easily to reproduce with `refresh()`. >> Turns out I was right, it could already happen before, without `refresh` >> involved. >> So I wrote a second test for that scenario. >> >> Both tests (and the bug of course) are fixed when we always clear the index >> of the cells that will land in the pile and to be removed from the viewport. >> This ensures that we will always update a cell we use from the pile. >> >> The reuse of the cells when scrolling a bit is unaffected. >> Example: When we show 10 cells, scroll a bit down so that the first 3 items >> are not visible anymore, we will still reuse the cells with index 3 - 9, and >> update the other 3 cells with their new index 10,11,12. Nothing changed >> here. If we now change the height so that cell index 12 is not visible >> anymore, it will land in the pile and will get the index `-1`. >> >> --- >> >> Another advantage: Cells in the pile will not have a reference to their item >> anymore. Cells can be in the pile forever, e.g. because the user change the >> height of the application. >> >> This might come in handy for >> [JDK-8390878](https://bugs.openjdk.org/browse/JDK-8390878) as well. >> >> --- >> >> Note: I still decided to also the `resetIndex` call for the pile in my >> second commit. It is not needed to fix the bug, but it makes sense. When we >> are rebuilding the `VirtualFlow`, we can and should not trust any cell. It >> therefore makes sense to reset the index pf pile cells here as well. >> >> --------- >> - [x] I confirm that I make this contribution in accordance with the >> [OpenJDK Interim AI Policy](https://openjdk.org/legal/ai). > > modules/javafx.controls/src/main/java/javafx/scene/control/skin/VirtualFlow.java > line 2857: > >> 2855: cell.setVisible(false); >> 2856: if (cell.getIndex() != -1) { >> 2857: cell.updateIndex(-1); > > we need to check whether the focus has moved elsewhere, for example in an > `onEDitCancel` handler, otherwise L2870 will steal the focus back from where > it was moved by the handler. I do had already nearly a year ago another idea how to solve this. Because this is all but not good. I will push a change where I would like to know your opinion. ------------- PR Review Comment: https://git.openjdk.org/jfx/pull/2347#discussion_r4298249888
