--- Comment #14 from Andreas L. Delmelle <>  2009-05-29 
12:05:18 PST ---
(In reply to comment #13)

Hi Vincent,

> The idea was that a penalty of value infinite and class EN_PAGE become a
> penalty of value 0 when considering a column break.

Aaah... of course, now I get it.

The general idea is obviously:
keep-*.within-line => infinite penalty of class EN_LINE, should result in the
break never being considered as a legal break.
keep-*.within-column => infinite penalty of class EN_COLUMN, should result in
the break being considered as a legal line-break, but not a page- or
keep-*.within-page => infinite penalty of class EN_PAGE, should result in the
break being considered as line- or column-break, but not as a page-break

> Like I said in my first comment this is very hacky and needs to be improved. 
> All the more than the actual value may not always be zero (e.g. when breaking 
> the 
> column at a hyphenated word).

Indeed. On top of that, there are the 'terminating' sequences of
penalty-glue-penalty that we use for the filler space after the last break.
Since the breakClass for the first penalty in that sequence remains unset (=
-1), it is also considered as a legal break (getBreakClass() != EN_LINE), while
it is precisely meant to prevent any break before the filler glue.

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

Reply via email to