Le 25/11/2019 à 11:40, Jonas Meurer a écrit : > Hi Emmanuel, > > Emmanuel Kasper: >> Le 23/11/2019 à 14:17, Jonas Meurer a écrit : >>> Package: cloud.debian.org >>> Severity: important >>> >>> Hello, >>> >>> for some reason, boxes for providers 'libvirt' and 'virtualbox' are missing >>> for >>> buster64 tag 'v10.1.0' from https://app.vagrantup.com/debian/boxes/buster64. >>> >>> Unfortunately, this breaks inital setup of debian/buster64 boxes with those >>> providers, as many packages from the 10.0.0 apt sources list caches don't >>> exist >>> in the archives any longer. >> >> Thanks you for your interest for the Debian Vagrant Boxes. >> Unfortunately we don't have point releases for Vagrant Boxes for Buster, >> see here for the rationale: >> https://lists.debian.org/debian-cloud/2019/09/msg00041.html > > Thanks for the pointer. I understand your rationale, but ... > >> The package cache for Base Boxes is anyway always updated, since we >> don't rebuild the boxes after every security upgrade which has been >> pushed to the archive. >> >> Concerning the package cache, what prevents you from refreshing it >> before installing new packages ? > > The problem seems to be, that Vagrant itsels doesn't update the package > cache before provisioning.
For virtualbox, vagrant per default tries to > install the VirtualBox Guest Additions. Does it ? Using the Debian package and for what I remember from Windows from two years ago, I don't remember pristine Vagrant trying to install the guest additions by itself ( could have changed though ...) Are you using the vagrant-vbguest plugin ? If that's the case, you could then use the Contrib Buster 64 boxes which has the guest additions already installed. https://app.vagrantup.com/debian/boxes/contrib-buster64 This fails if the package cache > wasn't updated in the meantime, as the referenced kernel package doesn't > exist any longer. > Also, isn't it that between point releases no packages are removed from > the achives? Therefore it should be less likely that outdated cache > causes errors, right? Less likely to occur, yes, but you still have the base problem -- You know an upstream is nice when they even accept m68k patches. - John Paul Adrian Glaubitz, Debian OpenJDK maintainer