Hi Felipe, On Sat, Apr 22, 2017 at 04:09:07PM -0300, Felipe Sateler wrote: > On Sat, Apr 22, 2017 at 9:45 AM, Guido Günther <a...@sigxcpu.org> wrote: > > Hi Felipe, > > On Tue, Oct 27, 2009 at 12:34:26PM -0300, Felipe Sateler wrote: > >> Package: git-buildpackage > >> Version: 0.4.59 > >> Severity: wishlist > >> > >> A pretty common workflow is the following: > >> > >> 1- Work on the package > >> 2- Upload > >> 3- Import new upstream version > >> 4- Create changelog entry for local testing > >> 5- Some more work > >> 6- Upload > >> > >> This works like a charm with --auto. However, when steps 3 and 5 are in > >> another order (eg, you fixed something and then updated the new upstream > >> version), git-dch misses all the entries from before the new changelog > >> entry. A way to pick up all the changes since the last released version > >> would be gratly appreciated. > >> I'm thinking this can be done by making git-dch ignore changelog > >> versions marked as UNRELEASED to guess the version (and thus having to > >> use -s), but that would mean having to parse debian/changelog instead of > >> using dpkg-parsechangelog. > > > > Sorry for picking this up _that_ late. I wonder if there really is > > s.th. to fix here. If you create a changelog entry (4.) afer doing local > > modifications you ought to do so with gbp-dch. If you don't, you're > > basically telling gbp-dch: "I want these entries ignored". If you create > > the changelog entry for local testing using "gbp-dch --auto" everything > > is fine. > > So you suggest using `gbp dch -a -N $new_upstream_version-1` when > importing a new upstream version? If that works, it could be useful. > Even better if we could tell gbp-import-orig to run it for us :)
You can even drop the -N part. gbp dch --auto is enough, it will pick up the new upstream version automatically. You can also have gbp import orig run it, e.g.: $ gbp config import-orig.postimport import-orig.postimport=gbp dch -S -a --debian-branch=$GBP_BRANCH && git commit --amend -C@{0} debian/changelog > > We could do as you suggested and ignore entries marked UNRELEASED but > > I'd rather have people use "gbp-dch" for the local testing entries as > > well or is that unreasonable? > > I don't think it is unreasonable. This is certainly a minor issue. Given the above I'm inclined to close this then or mark as wontfix. Cheers, -- Guido