[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: autopkgtest results influencing migration from unstable to testing



On 05.06.2018 22:54, Paul Gevers wrote:
Hi all,

On 06-06-18 08:52, Simon McVittie wrote:
On Wed, 06 Jun 2018 at 10:28:53 +0530, Pirate Praveen wrote:
ruby-state-machines and ruby-state-machines-activemodel should go
together and even when autopkgtest for the version is unstable passed,
instead of reducing the age, it is considered a regression because
autopkgtest for the version in testing failed and age is increased.

Isn't this a sign that ruby-state-machines has broken API, and should gain
"Breaks: ruby-state-machines-activemodel (<< 0.5)" or similar before it
migrates to testing? If so, then the autopkgtest is doing its job: it
has found a bug.

I fully agree. In the past this hasn't been such an issue because
typically timing would be all right, but...

If packages in unstable "should go together" then you need to tell
apt that, otherwise it will assume that arbitrary partial upgrades are
a valid solution. Testing migration uses the same metadata.

it is fragile to not document it. Autopkgtest will do the right thing if
you tell apt how to handle the dependencies. Indeed, the last couple of
months multiple packages needed a "Breaks" statement and everything was
all right afterwards.

agreed in principle, however it this really feasible? Usually this will result in re-uploads until you catch all these breaks.


Reply to: