Ok so every thing makes sense then?

Le jeu. 8 sept. 2016 22:18, Neil Gilmore <[email protected]> a écrit :

> Hi Pierre,
>
> (trimmed a bit)
>
> On 9/8/2016 3:07 PM, Pierre Tardy wrote:
> >
> > In this case I would expect two things
> >
> > If there are no connected workers for that build the build should stay
> > queued forever
> >
>
> It would, except that I've slightly modified the code to not consider
> revision when collapsing requests (we always want the most recent). So a
> build request gets swallowed when a new one is scheduled. I can see the
> most recent request in the queue.
>
> > If a worker is trying to connect with a spelling mistake on its name
> > the master should display a log saying that they is wrong
> > authentication for the worker and the worker should also constantly
> > complain about a authentication denied or something
> >
>
> To the master, it's not a spelling error (I've seen enough of those
> errors elsewhere, usually when I had to update a worker and botched the
> workername). The master thinks everything is fine. It just has an entry
> in its c['workers'] that no worker process will ever match.
>
> There's no worker process (by that name) to complain.  And no worker
> ever tries to attach by that name.
>
> Thanks!
>
> Neil Gilmore
> grammatech.com
>
_______________________________________________
users mailing list
[email protected]
https://lists.buildbot.net/mailman/listinfo/users

Reply via email to