Dmitry Smirnov wrote: > On Tuesday, 23 August 2016 8:51:23 PM AEST Adam D. Barratt wrote: > > That's not an excuse for causing disruption in unstable. > > I'm not sure when it is OK to cause disruption in unstable. For example > uploading new GCC seems to cause a lot of problems despite attempts to > mitigate FTBFS.
It's a very easy rule for protobuf, since protobuf has a non-trivial set of reverse build-dependencies: every ABI bump for protobuf needs a corresponding, coordinated ABI transition. For previous protobuf transitions (2.5.0, 2.6.0), please review #726165 and #760343. It's not as simple as just uploading a new release to unstable. Probably it should have been uploaded to experimental first, to check that the package would build and pass its test suite on all architectures. (E.g., see #572923 for an example of architecture-specific breakage in protobuf.) > Also do you have a clue why protobuf FTBFS on build servers? I'm unable to > reproduce the problem... I built it on amd64 in an up-to-date sid pbuilder chroot and it failed in the same manner as it did on all the buildd's. -- Robert Edmonds edmo...@debian.org