Hi Andreas, Andreas Delmelle wrote: > On 23 Jun 2009, at 17:50, vhennebert apache org wrote: > >> Author: vhennebert >> Date: Tue Jun 23 15:50:15 2009 >> New Revision: 787733 >> >> URL: http://svn.apache.org/viewvc?rev=787733&view=rev >> Log: >> Code clean-up >> > <snip /> >> http://svn.apache.org/viewvc/xmlgraphics/fop/branches/Temp_ChangingIPDHack/src/java/org/apache/fop/layoutmgr/FlowLayoutManager.java?rev=787733&r1=787732&r2=787733&view=diff > <snip /> >> - if (!(curLM instanceof WrapperLayoutManager) >> - && curLM instanceof InlineLevelLayoutManager) { >> - log.error("inline area not allowed under flow - >> ignoring"); >> - curLM.setFinished(true); >> - continue; >> - } > > This may be too much cleanup. I'm not entirely certain, but the > 'continue' statement is meant to prevent this condition from crashing > FOP on something we can perfectly recover from... (IIRC, without this > check, we would end up with a ClassCastException when adding the areas) > > Admitted, this event would better be routed through the event mechanism, > so users can decide for themselves. As long as that has not been done... > > Maybe nothing, since the check will always return false (should be > caught during FO tree validation, but then there's relaxed validation...).
o.a.fop.fo.pagination.Flow explicitly checks that every child element is a block-level element, even in relaxed validation mode. Also, after replacing the log with a ‘throw new IllegalStateException()’ the whole test suite ran without any problem. So I considered it safe to remove. > All the other changes receive my blessing! :-) Thanks for double-checking! Vincent
