Tuesday, 11 December 2018

Re: autopkgtest verision deployed on production infrastructure updated

On Tue, Dec 11, 2018 at 12:16:26PM -0200, Andreas Hasenack wrote:
> On Mon, Dec 10, 2018 at 8:11 AM Iain Lane <laney@ubuntu.com> wrote:
> >
> > +skippable
> > + The test might need to be skipped for reasons that cannot be
> > + described by an existing restriction such as isolation-machine or
> > + breaks-testbed, but must instead be detected at runtime. If the
> > + test exits with status 77 (a convention borrowed from Automake), it
> > + will be treated as though it had been skipped. If it exits with any
> > + other status, its success or failure will be derived from the exit
> > + status and stderr as usual. Test authors must be careful to ensure
> > + that ``skippable`` tests never exit with status 77 for reasons that
> > + should be treated as a failure.
> >
> Is it ok to use skippable for arches where the test is known to not
> work? AFAIK there is no arch restriction support in d/t/control
> I would then mark the test as skippable, detect the arch at runtime,
> and if it's one we know it won't work, exit 77. I understand care
> must be taken to check the test and make sure it doesn't exit 77 for
> other reasons in that case.

That sounds fine, but consider first if you can encode *what* about the
system makes the test fail, if that can be done in a better way than the
architecture itself. You might be able to use skip-not-installable if
this is expressable via dependencies.


Iain Lane [ iain@orangesquash.org.uk ]
Debian Developer [ laney@debian.org ]
Ubuntu Developer [ laney@ubuntu.com ]