On 21/08/14 11:42, Nate Finch wrote:
> The point of the test is that developers *don't* run godeps. That's the
Well they *should*. If they don't why not? Is it not documented well enough?
It's explicitly mentioned in the CONTRIBUTING doc.
> problem. The original email noted that our dependencies.tsv in master was
> wrong. It's not the first time this has happened. I've also seen many
> developers wonder why things were broken when it was just godeps needing to
> be run.
>
There's lots of ways around that. If they use the Makefile for example, make
build/install etc all depend on running godeps.
If dependnecies.tsv in master was wrong, that's an issue that needs to be fixed
in the landing tests and/or CI.
> The only time the test is bad is when you're working on a copy of an
> external repo.... And in that case, the failing test is a nice reminder
> that you need to update dependencies.tsv for your change to take effect in
> core.
>
And it now also fails for both me and Tim with the errors I posted earlier
dependencies_test.go:70:
c.Fatal(string(out))
... Error: godeps: no version control system found for
"/usr/lib/go/src/pkg/bufio"
godeps: no version control system found for "/usr/lib/go/src/pkg/bytes"
I and others agree with the intent of the change but not the implementation, nor
the impact it has on developers. Perhaps a change of this nature might have been
raised on the list before being committed to ensure there was broad agreement it
was the right thing to do, and didn't adversely impact things.
As it stands, right now, neither Tim, nor myself have passing tests. And
possibly others as they pull trunk.
I think this needs to be reverted until the issues are sorted out and we agree
on the best way forward.
--
Juju-dev mailing list
Juju-dev@lists.ubuntu.com
Modify settings or unsubscribe at:
https://lists.ubuntu.com/mailman/listinfo/juju-dev
