Should we MFC tests into stable/10?

[Garrett Cooper]

> On Mar 6, 2014, at 6:26, Julio Merino <jmmv at freebsd.org> wrote:
> 
> Hello all,
> 
> Keeping the testing infrastructure in sync between current and
> stable/10 is, in my opinion, a worthy goal. For that reason, I have
> (finally) pulled up a bunch of related pending changes into the branch
> today and will continue to do so for upcoming improvements.
> 
> However, I would like to gather your opinion on what to do about the
> tests themselves.
> 
> Do you think it's worth keeping the tests between current and
> stable/10 in sync wherever possible? Because we have barely just
> started adding tests, this will certainly involve quite a bit of churn
> in MFCs -- but that's probably not a big deal. The tricky^Winteresting
> cases will come when tests start failing in only one of the two
> branches :-P
> 
> My opinion is now leaning towards merging everything where it makes
> sense.  What's yours?

You're probably going to run into similar problems that devs run into when MFCing code.

It depends on whether or not the tests require a specific piece of build/test infrastructure, and whether or not the tests exercise a new feature or fix/change behavior.

The former case is trickier as there will be breaking points for things in the future, but the latter case is pretty straightforward (devs should be merging tests back with code changes for new features/behavior).

I vote yes for stable/10 now, but I would be wary of future merges as head and stable/(head-1) diverge, and similarly stable/(head-2).