[Gluster-devel] With what code/tree should the regression tests be run?

[Niels de Vos]

On Sat, Nov 15, 2014 at 11:46:46AM +0530, RAGHAVENDRA TALUR wrote:
> I could not reply to this thread yesterday.
> Earlier, test system used to take care of the rebase.
> However, it worked here only after I did a manual
> rebase on my machine and updated gerrit.

Indeed, and I have noticed that too. I remember being able to just
resubmit the failed regression tests, and the change would be rebased on
top of HEAD of the branch. To me, that also looks a safer way of going
things, (relatively) old patches can be tested easier that way.

I suspect that this change came with the triggered regression testing,
but I have not checked that. Do others have an opionion on with what
code the regression tests should be executed? Currently I am aware of
two possible ways:

a. checkout the HEAD of the branch and cherry-pick the change on top
b. checkout the change (use the tree from the time the change was
   posted)

(b) seems to be the current way, my preference would go to (a).

Please voice your opinions. Thanks,
Niels