On Fri, 2017-01-06 at 23:38 +0100, Andreas Tille wrote: > Hi Gianfranco, > > On Fri, Jan 06, 2017 at 09:33:31PM +0000, Gianfranco Costamagna wrote: > > > > I wonder whether you want to upload this version of h5py to unstable > > > > (rather than to experimental) since this seems to be a transition > > > > and we are in transition freeze. > > > > > > > > > > What do you mean by transition ? h5py has no abi and the changelog does > > > not mention any API breakage. Unless you had something else in mind ? > > > > I also looked at the diff, and I didn't find ABI breakage... > > did you see something I missed? > > I mean that > > $ apt-cache rdepends python-h5py | wc -l > 34 > > packages depend from this package and I was burned by a similar issue > when python-numpy was breaking one of my packages and other Debian > Science members packages as well. So did you tested whether all those > rdepends will work with the new version?
Oh, I see what you are referring to. I am afraid I have not, and I am putting my faith on upstream staying as conservative as they have been with their releases since version 2.4. I reckon a transition has never been needed for h5py since I took over its maintenance, and I am crossing my fingers this won't be the case for its 34 rdepends now (compared to 403 for numpy nonetheless). Ghis