[Patch] scrollbars

[Germain Garand]

Le Vendredi 30 Juillet 2004 08:57, Stephan Kulow a écrit :
> Because the regression testing is there to test for regressions. This very
> patch caused regressions and Dirk wanted to have a look.

I think I have something... I don't know if maybe Dirk could review it? See 
attached patch.

It appears our scrollbar problems stem from the fact that the view's content 
is always sized to at least the extent of the viewport.
Hence when a scrollbar is scheduled for showing, the space it takes on the 
viewport immediately makes the other scrollbar appear in the same go (from 
within QScrollView::updateScrollbars) because there is (artificially) no room 
for it in the content.

The same goes for shrinking the view when resizing: at this point, the 
viewport is always smaller than the content, so scrollbars tend to appear 
(#84798, also in the scope of the patch).

Unfortunately, as I get rid of this part:
-        // be optimistic and say that we never need a vertical
-        // scroll bar. fixes ugly cyclic recalculation chains
-        // with QScrollView.
-        QSize s = m_view->viewportSize(m_view->contentsWidth(),
-                                       0);
the [R] tree is different on about 500 testcases (the canvas width is now 
correct - i.e, reflects the vertical scrollbar status).

As for painting, there are about 20 scrollbar differences, and they look like 
fixes to me (e.g mozilla/dom/dom-html/hquo002.html didn't have a horizontal 
scrollbar at png generation time, now gets one)

Likewise, relevant testcases from #61730 look fine (#59,#65,#68,#72) ;
#69613 does not regress.

Greetings,
Germain
 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: scrollbars.diff
Type: text/x-diff
Size: 4088 bytes
Desc: not available
URL: <https://mail.kde.org/mailman/private/kfm-devel/attachments/20040803/404f6a79/attachment.diff>