On Mon, 2018-03-26 at 10:39 -0700, Adam Williamson wrote:
On Mon, 2018-03-26 at 17:06 +0000, Alessio Ciregia wrote:
> On Mon, Mar 26, 2018, 5:21 PM pmkellly(a)frontier.com <pmkellly(a)frontier.com>
> wrote:
>
> > Since the Gnome 3.28.0 test object for today was delivered as
> > Fedora-Workstation-Live-x86_64-28-20180324.n.0.iso, I went ahead and did
> > the rest of the testing. I was surprised when I could not report the
> > results. relval said there was no test event for 20180324.n.0. Did I do
> > something wrong
>
>
>
> Hello. Maybe someone else can explain better than me.
> However, every day one or more composes are produced (let's call them
> nightly builds) but you can report the results (directly on the wiki or via
> relval) only on composes nominated for test (test event). Such nomination
> is made automatically based on some heuristic rules.
> In fact, if you visit the wiki page containing the results [1] you are not
> supposed to find the latest compose, but instead the last nominated for
> testing.
> I think you can find more info here [2].
>
> I hope I didn't wrote anything wrong.
This is exactly correct, yes. A compose is attempted for Rawhide every
single day (at least), and a compose is attempted for Branched every
day that Branched exists. However, we do not create validation events
for every single compose, so as not to overwhelm people with events
(and the wiki with result pages).
The rules for when an event is created are a bit complex, but it's more
or less this:
1. We *never* automatically create events closer than 3 days together
2. If it's been 14 days since the last event, there will *definitely*
be an event for the next successful compose
3. If a compose succeeds and it's been more than 3 days but less than
14 days since the last event, then a new event will be created if any
of a hand-maintained list of 'important' packages changed since the
last event. If not, there won't be an event.
So yep, you can't expect there to be an event for every compose, at
least the way we do things at present.
Oh, forgot one very important rule. 1 isn't quite true, in the above
list. Proper list:
1. An event is *always* created for any successful candidate compose
(the ones we manually request from releng, as compared to the automatic
'nightly' composes...RCs are candidate composes).
1. We *never* automatically create an event for a new nightly compose
if it's been less than 3 days since the previous event.
2. If it's been 14 days since the last event, we will *definitely*
automatically create an event for the next successful compose.
3. If a compose succeeds and it's been more than 3 days but less than
14 days since the last event, then a new event will be created if any
of a hand-maintained list of 'important' packages changed since the
last event. If not, there won't be an event.
--
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | XMPP: adamw AT happyassassin . net
http://www.happyassassin.net