Hi,
On Sun, 10 Nov 2019 12:44:26 +0100 Matthias Klose <[email protected]> wrote:
britney doesn't run autopkg tests for binNMUs. E.g. for a library transition,
britney only runs the tests triggered by the library package, it doesn't run the
autopkg tests for all the packages built with the new library version. Things
known not to be catched are at least:
* A library rebuild causes an ABI change in another library. E.g. when
boost is rebuilt with a new version of icu, it changes ABI (that seems
to be not the case in recent boost versions).
However this kind of thing is currently not detected.
* binNMUs picking up unrelated changes, and failing. E.g. dh-python now
generates dependencies on python2 instead of python, but the autopkg
tests still call python.
I had not acted on this bug report for a very long time, with the argument that we don't want to throw away smooth-updates behavior by blocking transitions. But we don't need to block the library from migrating if it's smooth-updateable, what we want to block is the rebuilds if they cause regressions.
It turns out that adding that support is not very complicated as by adding a apply_srcarch_policy_impl method to the AutopkgtestPolicy already gets is called at exactly the right moment. All we need to do is add the right checks that are different than for source uploads.
I've implemented what I believe is what we want, but lets see if people agree that. See also MR123
Now, I wouldn't deploy this without discussion, because I think visibility of the results isn't great. The blocks will show up on the global excuses page [1], is that enough for tracking migration of binNMU's and their autopkgtest results?
Paul [1] https://release.debian.org/britney/update_excuses.html
OpenPGP_signature.asc
Description: OpenPGP digital signature
