[Libreoffice-qa] db test VM

[Yifan Jiang]

Hi Petr/all,

On Thu, Oct 27, 2011 at 09:08:11PM +0200, Petr Mladek wrote:
> > 	I (for one) was hoping that with openQA we could do a similar level of
> > UI testing, but based on what actually shows up rendered on the screen
> > (OCR'd) and that this might be more robust, and provide a fresh start
> > here that doesn't have testtool's problems.
> 
> Yi Fan did some investigation. It seems that openQA does not support any
> OCR. It probably just compares md5-sums of screenshot of the whole
> screen. I am afraid that this approach is not usable for LO because it
> is hard to maintain. Any different pixel will invalidate the test.
> 
> Reference:
> http://wiki.documentfoundation.org/QA/Testing/Subsequenttests

Just now, Berhnard and Dominik (Great thanks!) offered me an hour to show
things about OpenQA. The most impressive thing is, besides whole screen
md5sum, OpenQA also has a mechanism to 'imagegrep' if a small piece of image
is inside a whole screenshot.

For example:

http://openqa.opensuse.org/viewimg/opensuse/testresults/openSUSE-DVD-i586-Build0002-kerneldevel/ooffice-2.png

Notice the small pictures gallery below the big pictures, they are reference
images for the testing screenshot. If clicking the small scissor icon on the
top left of the screen, people can creat a small piece of the screen and the
system verify tests by the 'imagegrep' method. :)

Best wishes,
Yifan