Control: tags -1 - moreinfo Hi Adam,
Sorry for having been kept away from this bug for the last few weeks; life happened :( On Thu, Jun 02, 2016 at 06:48:41AM +0200, Adam Borowski wrote: > control: tags -1 moreinfo > > On Thu, Jun 02, 2016 at 01:31:02AM +0200, Nicolas Braud-Santoni wrote: > > > > The dsc and build artifacts can be found in > > https://nicolas.braud-santoni.eu/tmp/deb/ > > The dsc there doesn't unpack -- filenames nor checksums don't match. I'm not entirely sure what you mean there about a filename mismatch. In anycase, it turns out I was building on a machine with faulty memory, so the checksums mismatch might have been caused by this. > Only one of four commits atop of debian/1.0.0-1 looks relevant: > 40d4511 has the following changes: > + updated build-dependencies (ie, the meat of the NMU) > + standards version bump > + a changelog entry (with an error: you need # after "Closes: " before the > number) > > e12e613 adds an UNRELEASED changelog entry and some gbp junk. Indeed. I failed to clean this up while removing so-far-unreleased versions. > 3d0722a tries to invent a way of generating "orig" tarballs. > + this is bad as those don't match upstream. Besides details like losing > timestamps or degrading xz to gz, you really should use what upstream > provides, unless this is impossible because of git-only releases or a > need to repack. > + not using upstream tarballs ie especially bad as these are signed Yes, I didn't realise that the upstream packages were not from https://github.com/Yubico/libu2f-host/releases but from https://developers.yubico.com/libu2f-host/Releases/ (The rule that was added did reproduce the packages in https://github.com/Yubico/libu2f-host/releases) > 4e6802c tries to ignore changes in a generated file instead of properly > cleaning it (deletion would suffice) > > On the other hand, something in clean-up after build does happen to break > the package. If you try to package the source from a fresh repository, it > succeeds. After a build, newly repacked source won't build anymore: I'm not entirely sure what goes wrong there. Is it even within the scope of this NMU, given that it is a pre-existing problem unrelated to the RC bug I am fixing? I uploaded a new set of files to the same directory: https://nicolas.braud-santoni.eu/tmp/deb/ Note that the version number changed: it was following a previous UNRELEASED version, which could have led to it being preferred to a later release from the maintainer. Best, nicoo PS: Sorry for the double-sending: the first mail was sent to you but not to the BTS...
