Axel Dörfler wrote:
Adi Oanca <adioanca@xxxxxxxxx> wrote:
BTW, I'm pretty sure I know why BWindow's port gets filled up when using the new code and a B_VIEW_MOVED/RESIZED message per view. It's because in ServerWindow object we lock the RootLayer object for every operation that we need to do. Getting the view's coordinates following a BView::Bounds(), does not require RootLayer locking.
Since when? AFAICT it still needs a read-only lock (just not a write lock); the view coordinates can be changed by the root layer thread, so they are better read atomically.
Yes, true. I'll use MultiLock.