> On Sep 5, 2016, at 11:32 PM, Ajay Yadava <[email protected]> wrote:
> However, because of some project specific reasons community wants to run
> all tests to be sure of the patch. This results to another oddity that we
> don't want to run tests before the patch is applied.
Unit tests aren't run prior to the patch. The other tests are so that
a "before" status can be created. Without that "before" status, things are
likely going to be a bit wonky. Some plugins don't really work very well (hai
findbugs) and those that do will list *all* of the problems, not just the ones
caused by the patch. It's going to be a bit like running qbt against a patched
source tree.
> I am planning to avoid
> running tests when there are no java changes and this should help in many
> scenarios. I will create a profile accordingly.
Yetus already avoids running a lot of unnecessary tests by each plug-in
checking what got changed and whether that change should cause an execution.
You can tune it a bit via a custom personality_file_tests function. (See
hadoop's, which is pretty massive.)
