Hi, Quoting Samuel Thibault (2016-09-22 13:39:06) > Johannes Schauer, on Thu 22 Sep 2016 13:35:23 +0200, wrote: > > Thus, I do not think that following policy and only putting package names of > > packages that are actually produced into the Binary field of the .changes > > file > > should pose any problem. > > AIUI it does problem with packages built on some archs and not others: > say I have a packages foo which produces only the foo binary package on > amd64, but on arm it also produces a foo-arm package. When I make my > initial upload, ftp master will be notified about the foo binary package > only. When the arm build gets uploaded, the foo-arm package gets in the > way. > > That's why I was proposing to still include all potential packages, just > not the staged packages which don't appear in the unstaged build, and are > thus never to reach ftpmaster.
okay, then indeed, the first step should be to just include all packages that would build on any arch with no build profile active. At least we now know that doing so should not break anything, so we can go ahead and put such a change in dpkg-genchanges. This at least will fix this particular problem. For a check whether a new upload adds any packages that should go to NEW, the existing Package-List field should be considered instead of the Binaries field. As we see through this bug, the Binaries field is not expressive enough. Thanks! cheers, josch