On Sun, 6 Mar 2016, Yaroslav Halchenko wrote: > On Mon, 07 Mar 2016, Santiago Vila wrote: > > The following packages have unmet dependencies: > > sbuild-build-depends-pymvpa2-dummy : Depends: swig2.0 but it is not > > installable or > > swig (<= 3.0.5) but it is > > not going to be installed > > well -- swig2.0 was removed from testing but present in sid > > my guess is that pymvpa2 pkg would eventually be autoremoved from > testing, and there is nothing for me to do about it neither now nor > later to "close" this issue. > > FWIW -- can't use swig 3.x due to > https://github.com/swig/swig/issues/563 which could have been a > proper 'fix'. But as long as swig2.0 is part of the archive, issue is > not really an issue for me to tackle. So I will close it for now, feel > free to reopen if you feel it would be of some value (please describe of > which)
Ok, you reopened later but maybe I should explain a little bit: The testing distribution is supposed to be in an "always releaseable" state. Since packages in stable should never build-depend on packages not in stable, it follows that packages in testing should never build-depend on packages not in testing. So, even if we are not going to release stretch as stable today or tomorrow, this would still be a RC bug, i.e. a bug that must not be present in testing. The scripts that create the testing distribution from the "pool" ensure that dependencies are always met in testing, by construction, but I think they do not make the same check for build-depends (yet). So, if pymvpa2 is to be autoremoved from testing, it would be because of RC bugs like this one being open (which is why this bug is here to begin with), not because the testing scripts decide to remove it on their own. There are two common cases for bugs like this one: * The missing build depends package is in unstable but it has not entered testing yet. We sort of tolerate bugs like this one (maybe to the point of downgrading the bug to "important" to save some work), because we know it is a matter of time that they get fixed. * The missing build depends package was in testing but it was removed to not come back again. This is a RC bug in all cases. As I am building packages with "dpkg-buildpackage -A" in a random fashion, not because I use those packages myself, I don't have any familiarity with the affected packages, so I leave to the maintainer to decide which one of the two common cases is the one that actually happens. Thanks.
