We are seeing a problem occasionally where, when switching the data displayed in the grid, some of the prior rows continue to display (see attached). These rows seem to just hang there since their location is not necessarily immediately following a valid row. Which is one way we know these are not actual rows that are part of our dataset.
One way to make this happen is to go into insert mode on a cell and then click elsewhere (in our case it is in a tree control) to make the grid reload with new data that has fewer rows than it originally had. The hard part of this is that I have not been able to reproduce this in a sample application. Our actual application's use of the grid is quite rich with features (virtualization, attached properties, non-auto generated columns, asynch dataloading) so it is not easy to bring all this over to a sample. I do have my sample using the above except for the attached properties.
Has anyone seen anything like this? Any clues as to how to prevent? It doesn't just happen if you are in insert mode but that is a sure-fire way to make it happen. I've tried having the code call CancelEdit on the grid before it switches the data but that did not help.
I've attached 2 screen shots. One shows the phantom rows without having been in insert mode prior to capturing this screen. The other shows the case when the user was in insert mode. You can see the phantom row showing the insert indicator on the left. I've drawn a red box around these phantom rows in these screen shots.
Thank you, Valerie