Hi On 2020-11-02 19:18:45 +0100, Paul Gevers wrote: > HI Stéphane, > > On 02-11-2020 16:56, Stéphane Glondu wrote: > >> autopkgtest for ocaml-cairo2/0.6.1+dfsg-5: amd64: Regression ♻ (reference ♻), arm64: Regression ♻ (reference ♻), armhf: Regression ♻ (reference ♻), i386: Regression ♻ (reference ♻), ppc64el: Regression ♻ (reference ♻) > > > > However, I don't understand why it complains about > > ocaml-cairo2/0.6.1+dfsg-5 whereas version 0.6.1+dfsg-6 is fixed. Does > > something need to be done? > > Your symptoms describe a missing *versioned* Depends/Breaks or test > dependency. The migration software tries run tests with the minimum > changes. So, if nothing makes sure that the version from unstable is > needed, it will run with as much packages as possible from testing. What > we see in the logs is that it installs the test suite from testing. > Then, the test dependencies are installed and apt-get reports that it > can't install the required packages. It then falls back to enable > everything from unstable and then installation is possible. > > Do you confirm that the issue is only in the test? I think we can ignore > this issue then (or better, trigger the test manually with ocaml-cairo2 > from unstable. With the description above, do you understand why the > relations in the packages don't teach our migration software what to do? The test fails because it tries to build the examples contained in the source package. The version in testing, 0.6.1+dfsg-5, fails to build with ocaml 4.11.1, so that's expected. The issue was fixed in 0.6.1+dfsg-6. Since Depends and Breaks do not affect source packages, I don't think this is fixable in any of the involved packages other than by fixing the FTFBS bug. I think this issue can be ignored and I will add a hint. Cheers -- Sebastian Ramacher
Attachment:
signature.asc
Description: PGP signature