Run apprunner and choose "Edit Bookmarks" from the menu. The menubar changes to the bookmarks menu. However, neither the close command nor clicking the close box works to close the window. You have to use the debugger to kill the app.

assigning to danm as p1 for m3, because it sounds related to the other bookmark window bug you've got.

You can't close any window on the Mac. That's because there's no way to programmatically close a window. I suspect you can close a window on Windows or Linux only because the OS has already done so after you hit the "close" widget. We're not actually doing the close ourselves. The solution is to write a "close window" method, which, it turns out, is not trivial, though it's been on my plate for a while now.

Previous pontification not quite correct; it was actually a horrible, intertwined refcount problem. The window was never deleted because its refcount could never be zero. It's fixed at the current time. I make no promises about next week.

Earlier fix had bad side effects and was partially backed out. I just checked in a slightly different fix without those particular bad side effects. Consider this my re- opening and then re-resolving this bug.

elig, please check this with todays Mar16 build when available. Thanks!

I see this problem readily on the 3.15.99 build, but it's not reproducible for me on the 3.16.99 build. The bookmarks window (and all other windows) have been easily closeable. (Specifically, using the close widget; it doesn't work no matter what using the menu item, which I'm assuming is unrelated, since there's lots of menu item bustage.)

Reopening because the fix caused a regression in Viewer: the app crashes everytime a window is closed.

Clearing the resolution of Fixed

The regression introduced by this fix is described in bug #3900

Note: I don't think that we can choose to ignore the crash in Viewer because embedding apps are also concerned.

adding some comments ...

Hokay. Hacks rerouted through better code. I claim that once again the bug is fixed and further I assert that no nasty regressions result. There's still some refcount hell running amok, but the situation has been improved so much that this particular bug has once again evaporated.

Can't reproduce in Viewer or Apprunner on Mac OS 3.19.99 optimized build. Will mark as verified, but request that Pierre double-check with his blessing, too.

Moving all Apprunner bugs past and present to Other component temporarily whilst don and I set correct component. Apprunner component will be deleted/retired shortly.