[Qgis-developer] False failures in the testsuite

[Sandro Santilli]

Once again I'm facing testsuite failures from a branch with NOTHING
effectively changed. These are the changes in the branch:

  https://github.com/qgis/QGIS/pull/1729/files

Can you spot a real difference ?

And this is the corresponding travis build:

  https://travis-ci.org/qgis/QGIS/builds/43741928

Before orfeo dash was reset (does it happen daily?)
I could see one-pixel-border-line differences in some
labels test, which I've seen before while working on the
core rotation support.

As once again I cannot see what difference could my change introduce,
can it be that the tests are suffering from some code-independent
glitches making them unstable ? Could it maybe be threading ?
Or unclearn builds from travis ?

--strk; 

 ()  ASCII ribbon campaign  --  Keep it simple !
 /\  http://strk.keybit.net/rants/ascii_mails.txt