broken chart2_unoapi (was: Tinderbox failure, Linux-RHEL6-x86_64@14-with-check, MASTER, last success: 2012-03-22 05:28:28)

Hi all,

If somebody into chart and/or calc -- yes Markus, I'm looking at you :) 
-- could have a look at this failure, that would be highly appreciated. 
 Noel said he intends to look into this, too, but would take quite a 
while and he wouldn't mind a helping hand; and I would like to see this 
test go green again as soon as possible.

Thanks,
Stephan

-------- Original Message --------
Subject: Re: Tinderbox failure, Linux-RHEL6-x86_64@14-with-check, 
MASTER, last success: 2012-03-22 05:28:28
Date: Thu, 22 Mar 2012 16:36:06 +0100
From: Stephan Bergmann <sbergman@redhat.com>
To: Noel Grandin <noel@peralex.com>
CC: fridrich.strba@bluewin.ch <fridrich.strba@bluewin.ch>

On 03/22/2012 02:31 PM, stephan.bergmann.tinderbox@googlemail.com wrote:
> One of you broke the build of LibreOffice with your commit :-(
> Please commit and push a fix ASAP!
>
> Full log available at http://tinderbox.libreoffice.org/MASTER/status.html
>
> Tinderbox info:
>
>    Box name: Linux-RHEL6-x86_64@14-with-check
>    Branch: MASTER
>    "starttime": 1332409825
>    Machine: Linux desktop01 2.6.32-192.el6.x86_64 #1 SMP Tue Aug 23 14:34:08 EDT 2011 x86_64 x86_64 
> x86_64 GNU/Linux
>    Configured with: --disable-dependency-tracking
> --disable-zenity
> --enable-binfilter
> --enable-dbgutil
> --enable-symbols
> --enable-werror
> --with-max-jobs=2
> --with-num-cpus=2
> CC=/usr/bin/gcc
> CXX=/usr/bin/g++
>
> Commits since the last success:
>
> ==== core ====
>    a75fd2d  Look for properly registered windows sdks instead for random version numbers
>    7ec7f0d  MinGW apparently needs to be forced into __cdecl
>    ab2186d  MinGW: other DLLs must not pick sal's special __main
>    4fc8059  Fix SAL_CALL placement (so MinGW won't complain)
>    17c4bb3  Needless SAL_CALL
>    e1b4216  ChartTokenMap can contain null pointers
>    be79f65  Some changes in libcdr to appease valgrind
>    0589dd3  fix debug build
>    a603634  Fixed previous commit for --enable-dbgutil mode
>    e94c4ab  Conver SV VARARR to std::deque for sw module.
>    db3786e  Convert tools/table.hxx to std::map in Chart2PositionMap class
>    6af00cf  remove unnecessary use of local variable and auto_ptr
>    13bf197  if/else branches contain same code
>    4aa72e0  restructure insert code as a precursor to further simplification
>    3ffdb45  simplify code - remove unnecessary and complicated allocation
>    11ccf88  SVG import: keep aspect ratio when transforming viewbox to viewport
>    7d7e2c3  Convert tools/table.hxx to std::map in class ScHTMLLayoutParser
>    157b2fa  Convert tools/table.hxx usage to std::map in ScEEImport class
>    000bb6a  Convert tools/table.hxx usage to std::map in Basic IDE module
>    e55595d  maTransform had to go ( was always = identity)

With e1b4216 fixing the crash in chart2_unoapi test, chart2_unoapi still
fails (reproducibly, both for the tinderbox and for my local builds), at

> checking: [sch.ChXDiagram::com::sun::star::chart::BarDiagram] is iface: 
> [com.sun.star.beans.XPropertySet] testcode: [ifc.chart._BarDiagram]
> LOG> Setting Diagram type to BarDiagram
> LOG> Set it to 3D
> LOG> Execute: Vertical
> LOG> try to set:
> LOG> old = false
> LOG> new = true
> warn:legacy.osl:19200:1:/data/lo/core/svx/source/sdr/contact/viewcontactofe3dscene.cxx:387: No need 
> to get ViewInformation3D from an empty scene (!)
> warn:legacy.osl:19200:1:/data/lo/core/svx/source/sdr/contact/viewcontactofe3dscene.cxx:387: No need 
> to get ViewInformation3D from an empty scene (!)
> LOG> Property 'Vertical' OK
> LOG> result = true
> Method Vertical finished with state OK
> LOG> Vertical: PASSED.OK
>
> LOG> Execute: Deep
> LOG> property 'Deep' is not set but is not MAYBEVOID
> Method Deep finished with state FAILED
> LOG> Deep: PASSED.FAILED

which looks like it is likely due to db3786e, too.  I tracked it down to
chart::WrappedPropertySet::getPropertyValue ->
chart::wrapper::WrappedStackingProperty::getPropertyValue ->
chart::wrapper::WrappedStackingProperty::detectInnerValue ->
chart::DiagramHelper::getStackMode going into its two nested for loops
just once, calling getStackModeFromChartType, which calls
chart::ChartType::getDataSeries, which has an empty m_aDataSeries, so
getStackModeFromChartType does not go into its for loop, does not set
rbFound=true, so detectInnerValue returns false, so getPropertyValue
returns a void any.

Just why that m_aDataSeries is empty escapes me, though.  (To reproduce,
you can shrink chart2/qa/unoapi/sch.sce to just "-o sch.ChXDiagram".)

Stephan