Bug 325091

Comment 1 Ivan Furnadjiev 2010-09-13 05:31:09 EDT

Created attachment 178724 [details]
Screenshot

Reproducible with FF 3.6.9 on Windows and MAC.

Comment 2 Ivan Furnadjiev 2010-09-13 05:31:54 EDT

Created attachment 178725 [details]
Snippet to reproduce it

Comment 3 Tim Buschtoens 2010-10-13 05:33:42 EDT

Note that i was unable to reproduce this on my windows machine, but under Mac OS X (same FF version).

Comment 4 Tim Buschtoens 2010-10-13 06:53:41 EDT

There are more issues with the horizontal scrolling: Setting scrollLeft (via showColumn) with the first request or while the tree is not visible will have no effect or produce a JS error. Also, the columnHeader will not be scrolled correctly if it is made visible via setHeaderVisible after the Tree has been created and scrolled. 

I won't open new bugs for these, as this can all be fixed with one additional "onAppear" listener in Tree.js.

Note that these additional issues are bugs in the Tree.js, while the original issue with the wrong offset after reload is veryl likely a Firefox bug.

Comment 5 Tim Buschtoens 2010-10-13 10:07:10 EDT

Created attachment 180768 [details]
"improved" snipped

This is a "more risky" version of the snippet: It scrolls an invisible Table with an invisible header on the first request. (Doubleclick the shell to make the table visible and click the tree to make the header visible.) Only if both the clientArea and the header are scrolled to column 12 on appear does the tree work correctly.

Comment 6 Tim Buschtoens 2010-10-13 10:16:23 EDT

Note: I found a similar bug in IE that lets this test fail:
      var testUtil = org.eclipse.rwt.test.fixture.TestUtil;
      var tree = this._createDefaultTree();
      tree.setItemMetrics( 0, 0, 1000, 0, 0, 0, 500 );
      testUtil.flush();
      tree.hide();
      testUtil.flush();
      tree.setScrollLeft( 160 );
      tree.show();
      assertEquals( 160, tree._horzScrollBar.getValue() );
      assertEquals( 160, tree._clientArea.getScrollLeft() );      
      tree.destroy();

The second assertEquals gets a "0" not "160". However, the slightest changes (like making the header visible) to the test and it is green again. I was also unable to reproduce this problem in an actual RAP application so i will comment the test in TreeTest.js for now.

Comment 7 Tim Buschtoens 2010-10-13 10:25:41 EDT

Fixed in CVS HEAD.

Fixed in Tree.js and added tests to TreeTest.js.

Comment 8 Ivan Furnadjiev 2010-10-13 13:28:15 EDT

Did a testing under FF 3.6.10 and IE8 (Windows 7)... works fine.