[haiku-appserver] weird R5 behaviour with regards to follow mode

Hi all,

the other day, I had a weird problem in one of my apps with regards to view 
follow modes. I had a couple of views set to B_FOLLOW_NONE. Their parent 
view was set to B_FOLLOW_RIGHT | B_FOLLOW_BOTTOM. When the parent view was 
moved because its parent view was resized, the B_FOLLOW_NONE children 
actually stayed fixed with regards to window coordinates.
So I've written a small test app for the test environment that was supposed 
to demonstrate the behaviour. It turned out however, that R5 behaves 
inconsistently.

hierarchy:
View1 -> View2 (B_FOLLOW_RIGHT | B_FOLLOW_BOTTOM) -> View3 (B_FOLLOW_NONE)

Indeed, if View2 is *moved* within its parent because the parent is 
*resized*, View3 stay fixed (*moves* within View2 so that it stays fixed in 
window space). However, if View2 is moved programmatically then View3 moves 
along as if it had B_FOLLOW_LEFT | B_FOLLOW_TOP!!

In our app_server, B_FOLLOW_LEFT | B_FOLLOW_TOP is actually implicit. 
Children views *always* move along with their parent. The follow modes are 
only interesting with regards to parent views being resized. Since you 
cannot resize a views left or top side (that's considered moving the view 
plus resizing it), B_FOLLOW_LEFT and B_FOLLOW_TOP don't really have any 
meaning. I think our logic is consistent (the coordinate system is simply 
inherited).

I can not see the logic in what R5 is doing. If it worked in such a way 
that B_FOLLOW_NONE makes sure a view stays fixed in window space regardless 
of its parent, then it would make some sense. However, if the difference is 
the *reason* for the parents movement, then I consider it a bug.

Any comments (should B_FOLLOW_NONE make sure a view stays fixed in window 
space)?

Best regards,
-Stephan

Other related posts: