Issue 101169

Created attachment 61646 [details]
test case with several effects

I tested it on openSUSE 10.3-i586, KDE 3.5.10 with OpenOffice.org 3.0.1-de and
OpenOffice.org 3.1rc1-de (RPM) and can confirm the issue. I tried it out with
the animation turning (drehen) and slow. It worked on OOo 3.0.1. If I run the
test on 3.1rc1 the whole sentence was turned.

Reassigned to AF.

Add CCs.

Information about the animated text is taken from meta files.  Character and
word boundaries are encoded in comments.  These somehow got lost in the recent
transition to a primitive based drawing layer.

A simple fix for this is this:

Index: source/processor2d/vclmetafileprocessor2d.cxx
===================================================================
--- source/processor2d/vclmetafileprocessor2d.cxx       (revision 270468)
+++ source/processor2d/vclmetafileprocessor2d.cxx       (working copy)
@@ -924,7 +924,7 @@
                                        // restore DrawMode
                                       
mpOutputDevice->SetDrawMode(nOriginalDrawMode);

-                                       if(pTextDecoratedCandidate)
+                    //                                 if(pTextDecoratedCandidate)
                     {
                         // support for TEXT_ MetaFile actions only for
decorated texts
                         if(!mxBreakIterator.is())


With this fix the animation looks largely as before.  However, there is a long
delay between lines.

The long delay between lines is probably related to too many "words" that are
created with the "fix".  Here is the count of "words" for the first slide:

OOO300m15: 33
DEV300m45: 3
DEV300m45 with "fix": 89

@aw: Please take over.  You know the metafile generation from primitives much
better than I do.

I would suggest to concentrate on the question why pTextDecoratedCandidate is
null....

AW: Okay, taking a look. Comparing DEV300 m29 (last version before primitives)
with OOO310 m7. This shows that the preparation is pretty identical, but the
decision for pTextDecoratedCandidate has to be removed (as already stated).
The 2nd issue with the too many generated breaks is quickly found: The code in
drawinglayer/source/processor2d/vclmetafileprocessor2d.cxx uses rTxt =
rTextCandidate.getText() and rTxt.getLength() for the BreakIterator looping and
processing. This is generally wrong, since the TextSimplePortionPrimitive2D
works with String, Position and Length parameters; thus currently e.g. for a
text portion with start at 50 and length of 30, the whole string was tagged
again and again for each single portion. Corrected, that, works as expected.
Tesing a little more...

Created attachment 61702 [details]
Diff showing the fix

AW: Looks good, compared BugDoc animation completely with old behaviour, looks
identical now. Attached the fix.
It's not a dangerous one, it's very locally bounded in
drawinglayer/source/processor2d/vclmetafileprocessor2d.cxx and only used for
this MetaFile tag generation (one advantage of modular architectures with
primitives :-)).

I can confirm that the fix restores the original behavior.

AW: Changing target, adding to CWS aw072, targeted for OOo 3.1...

AW: Checked in, done.

AW->WG: Please verify. To do so, use the BugDoc. When starting the presentation,
word/character wise will work now as in OOo 3.0.

Verified in CWS under Linux and Windows XP.

Tested in master m11. Closed.