Control: tags -1 patch
Control: forwarded -1 
https://salsa.debian.org/ci-team/autopkgtest/merge_requests/12

Hi Martin,

On 17-05-18 11:20, Martin Pitt wrote:
> TBH I don't like this magic at all; it doesn't take into account virtual
> packages, version/architecture constraints, etc.

Hmm, indeed.

> It seems to only add
> this to the test dependencies if the package name actually exists (thus
> failing on unsatisfiable versions); how does this cause a failure if the
> package does /not/ exist (i. e. |apt-cache show| fails)?

It will typically fail during running the test and it will be unclear
why. What I try to get here is that it fails (most of the times) with a
clearer error message. See e.g.:
https://ci.debian.net/data/autopkgtest/testing/amd64/j/jupyter-sphinx-theme/248894/log.gz

> If a user installs a package with a non-available Recommends:, it will
> also not fail. Sure this is a packaging error, but one that britney
> should report and complain about, not autopkgtest.

The problem is that britney still triggers the test. Also, it may only
be a problem of the TEST dependencies. I don't think britney should
check those to be honest. So *if* we agree on that, it is up to
autopkgtest to try to create a error message that is as clear as possible.

Paul

Attachment: signature.asc
Description: OpenPGP digital signature

Reply via email to