Will put that into CWS thb11, @pl - care to have a quick look at the salbmp
change, not sure why this was missing in the first place?

Created attachment 57313 [details]
The patch

Fixed in CWS canvas06

@pl: please verify - build in vcl/test & run canvasbitmaptest; if that does not
turn up any errors, all is well (ok, all that's tested there is well...)

grmblfx

.

@pl: sorry for the noise. see above.

cleaned up dndtest.cxx so it compiles without warnings (wow that is old stuff !)

more seriously I got these warnings on Linux:

vcl/test/canvasbitmaptest.cxx: In member function `virtual
com::sun::star::uno::Sequence<com::sun::star::rendering::ARGBColor>
<unnamed>::TestBitmap::convertIntegerToPARGB(const
com::sun::star::uno::Sequence<sal_Int8>&)':
vcl/test/canvasbitmaptest.cxx:801: warning: passing `const double' for
converting 1 of `double vcl::unotools::toDoubleColor(sal_uInt8)'
vcl/test/canvasbitmaptest.cxx:802: warning: passing `double' for converting 1 of
`double vcl::unotools::toDoubleColor(sal_uInt8)'
vcl/test/canvasbitmaptest.cxx:803: warning: passing `double' for converting 1 of
`double vcl::unotools::toDoubleColor(sal_uInt8)'
vcl/test/canvasbitmaptest.cxx:804: warning: passing `double' for converting 1 of
`double vcl::unotools::toDoubleColor(sal_uInt8)'

These indeed seem to indicate problems since fAlpha is a cast sal_uInt8 itself,
so in the range from 0 to 255 -> overflow.

attaching a patch to solve that.

With that patch it passes, but only if I compile it with debug, else it crashes
somewhere in vcl while trying to access a service.

Created attachment 57976 [details]
fix warning / overflow

reopen

ah, good catch. sadly, gcc 4.2 is everything but -Werr clean on x64, so this one
escaped me. Will go for the crash now.

Hm. Fixed the warning (a bit differently) - but no crash here. From where do you
run the binary, I had to copy it into an installed office (openoffice3/program
dir), and run it with an LD_LIBRARY_PATH containing both ure & basis layer
(yeah, lame. but really no ambition to find a different way through the 3layer
OOo maze)?

@pl: could you try again, please?

verified in CWS canvas06

and no, I don't want to find my way through those layers either :-)

integrated, closing