On 2/6/19 4:17 PM, Adam Williamson wrote:
Rawhide isn't really utterly broken: a GTK+ change broke lots of
openQA
needles (screenshots) so the tests are failing. I'm in the middle of
re-taking all those now, and will re-run the tests on prod when it's
done, and send a new compose check report.
I am curious---if multiple tests fail this way, how do you determine
whether the changes are innocuous and re-taken screenshots can be used
from now on?
I assume there is no automatic criterion, or else you'd be using it for
the test, so you just have to eyebal all of them, right?