Issue 108052

Created attachment 66961 [details]
Impress Doc with animation that will play in 3.1

Reproducible with "Ooo 3.2.0 RC1 WIN XP DE-multilingual version German UI
activated [OOo320m8 (Build 9472)]"! On second slide item "b" ("Sermon usually
ends ..." will not be shown when playing presentation. Item is shown in MS PPT
Viewer and  with "2.4.1  Multilingual version English UI WIN XP:
[680m17(Build9310)]"

@kdvzhome2:
Pls. attach Source .odp from 3.0!
It would be useful if you could provide concrete information instead of a
general "does not work correctly". Please contribute information to other
problems if you saw them during your tests!
Before you file further issues or post again here, please read our guidelines on
<http://qa.openoffice.org/issue_handling/pre_submission.html> and  
<http://www.openoffice.org/bugs/bug_writing_guidelines.html>, then
contribute a clear step by step instruction containing all observations (error
messages ...), _every_key_press_and_every_mouse_click_ how to reproduce the
problem, and explain why you believe that your results are unexpected.
That means (for example): 
 do not write something like „I am not able to ...“, but
 „6. left mouse click on …
     expected: …, color of … changes, … 
     actual: no …., color remains white, no …

Screenshots might help to explain the problem.

Please always specify OS and Platform you used for your tests!

Reproducible. Reassigned.

@wg: What secret information do you have concerning reporters OOo version? ;-)

I believe version OOo 3.1 RC2 from report was correct, because that's the same
build as 3.1.1. I will set back Version because I can reproduce that problem
with "Ooo 3.1.1 WIN XP DE[OOO310m19 (Build 9420)]"!

Sorry, no secret information ;-)  I just set it to the newest version that I
could reproduce it in. Mostly a developer tells me: "what? found in version 3.1?
Sure it hasn't been fixed in the meantime?" I wanted to avoid this...

Looks like some internal data structure of the slide show is corrupted when the
last effect on the first slide is played (in makeTreeNode() in
slideshow/source/engine/shapes/drawshapesubsetting.hxx).
Needs more investigation.

The problem mentioned above is due to an uninitialized iterator (when
calculating the distance between two iterators, both have to point to the same
container, which here they did not.)

After fixing this I was able to reproduce the animation problem and find its
cause.  Only a solution is still elusive.

The animation problem can be seen on slide 2 of the attached bugdoc.  Text
paragraphs are animated to be show one after the other.  On this slide, however,
not all paragraphs are shown.

This is caused by the empty paragraphs in between.  In the internal
representation as drawing primitives these paragraphs are optimized away.  This
results in different indices for the remaining paragraphs then those used in the
animation descriptions.

Created attachment 75846 [details]
Proper initialization in DrawShapeSubsetting

@aw: Adding primitives and thus metafile actions for empty paragraphs would
probably fix this bug.  But maybe there is a better solution?

AW: I have a deva-vu here since I think this was already once coming up and had been fixed. After some debugging, I found the following line in impTextBreakupHandler::impFlushTextPortionPrimitivesToLinePrimitives():

// only create a line primitive when we had content; there is no need for
// empty line primitives (contrary to paragraphs, see below).

This was exactly the former fix. Thus, somehow this had been broken again. Base to have this working is that even for empty paragraphs, the DrawPortionHdl of the EditEngine has to be called. This is in impedit3.cxx in editengine module. Indeed, it gets no longer called with empty paragraphs because of this loop:

for ( sal_uInt16 y = pLine->GetStartPortion(); y <= pLine->GetEndPortion(); y++ )

OTOH the above text shows that it was called in the past. Unfortunately the move from svx content to own project editengine has nit taken over the direct history. It now gets not called since in the following switch the text portion kind is of type PORTIONKIND_TAB. That type supports no callbacks at strip at all. I'm really curious what might have changed to break this functionality again...

AW: OOps, copied the wrong comment. Here is the correct one:

// ALWAYS create a paragraph primitive, even when no content was added. This is done to
// have the correct paragraph count even with empty paragraphs. Those paragraphs will
// have an empty sub-PrimitiveSequence.

Looking what might have changed in EditEngine...

Created attachment 76175 [details]
AW: Supposed patch for this task

AW: Works, checked in, done.

AW->WG: Please review. Use the attached BugDor or alternatively create a presentation with a text frame with empty paragraphs (empty returns) and text animation.

Verified in CWS.