On 10/14/2014 04:08 PM, Markus Mohrhard wrote:
So it is a bit complicated. In contrast to draw/impress the xshape test in chart2 are not perfectly stable across machines. There are many factors that may result in differences between different computers. The reference machine for these tests was the @50 tinderbox which always produced the same results as my local Linux machine (which of course was helpful for me). Personally I regard all test failures of these tests a bit sceptical. I always check if there was a chart2 change that explains the test failure and if so regenerate the reference files (of course checking it is not a regression is still necessary). If there is none it is for sure a regression and the idea was that it is easier to find some of the problems introduced by commits to drawinglayer already through the automated tests. So in a perfect world these tests would be stable and everyone could run them, in reality it is best to run the only on a few machines that are known to work and maintained by people who can judge if a test failure is a regression or not.
now updated <http://cgit.freedesktop.org/libreoffice/core/commit/?id=f57a697fac4e2c885bfaf5add1dfe05ede552cd5> "Add clarification of --enable-chart-tests"
Stephan