Andreas L. Delmelle <> changed:

           What    |Removed                     |Added
  Attachment #23916|0                           |1
        is obsolete|                            |

--- Comment #38 from Andreas L. Delmelle <>  2009-07-15 
15:19:07 PST ---
Created an attachment (id=23990)
 --> (
Yet another update

Had hoped to get to this sooner, but it seems that at some point when getting
the existing testcases to succeed, I made a change that, again screwed up the
nested column/page keep test in the attachment. Took me a while to track it

The way the algorithm operates, after the changes in the previous patch, is
that in block-4 in the sample, we get the effect that the overflow condition
for the first page will be detected at the first break after the nested
block-4a (mentioned in comment #1 as something to watch out for). This means
that my devised strategy of keeping track of the keep-context would not work.
The keep context would switch again before we detect the overflow.
So I had the idea of making the skip dependent on one more factor: first
compute the difference. If that is negative, then we know that there will be an
unavoidable column-break somewhere before the current node, so we can already
trigger the overflow handling. Seems to work nicely, so far. Still have to
complete the testcases: the additional one present in the patch still needs
decent checks, and additional tests checking the behavior in tables and lists
would also be nice. Maybe it can be done by extending some of the existing
testcases. Getting into that right now.

Configure bugmail:
------- You are receiving this mail because: -------
You are the assignee for the bug.

Reply via email to