On 2015-05-06 10:44, Thorsten Glaser wrote: > [ with my m68k buildd maintainer and (ex-?) porter hat ] > > Aurelien Jarno dixit: > > >- debian-ports uses mini-dak instead of dak. It uses less resources and > > brings some features that are useful for new architectures like > > accepting binary uploads when it "improves" the version even if it is > > not the latest one or an "unreleased" suite for packages built from > > modified source (hence architecture specific). On the contrary its > > There’s two more bugs that *really* disturb porters’ work in it: > > • it is possible to do a binary upload of the *same* version of a pak‐ > kage that is currently in the archive, which breaks the package until > the next bigger upload fixes it: mini-dak serves the checksums from > one of the uploads but the .deb files from the other of the uploads > (this can happen in case of human errors, or caused by a w-b hiccup, > when a package was taken by someone (porter or buildd) and is in > Building state, then vanishes from the DB, then comes back) > > • (much worse) library transition old-version keeping is broken: > > suppose there is src:isl (= 0.12.2-2) building libisl10 in the > archive and built on some architecture; then, someone uploads > src:isl (= 0.14-2) which builds libisl13 instead; in dak (on the > main archive), the old libisl10 binary packages are kept in the > Packages.gz file until there is no dependency on it any more; > mini-dak just throws all NBS binary packages away immediately Then that's only one more, because it's exactly what I described in my mail. Aurelien -- Aurelien Jarno GPG: 4096R/1DDD8C9B aurelien@aurel32.net http://www.aurel32.net
Attachment:
signature.asc
Description: Digital signature