Hi, On Wed, Nov 15, 2017 at 03:02:33PM +0100, Víctor Cuadrado Juan wrote: > > > On 14/11/17 22:47, Guido Günther wrote: > > Hi, > > > > This wired and I wouldn't expect uncommitted changes but since I don't > > know about your setup and what you're doing (the upstream repo > > e.g. already has the version you're trying to import) you'd have to > > provide better instructions to reproduce and tell me what you actually > > think is wrong. > With git-buildpackage 0.8.12.2 in Stretch, if I take Debian's guitarix up > until 0.35.6, I can do > > gbp --import-orig --uscan --merge-mode=replace > > and it will correctly import the new 0.36.0 upstream version, merge it to > master > and preserve the already existing contents at debian/*. > > With git-buildpackage 0.9.2 (today in Testing and Unstable), doing the same > will overwrite the contents at debian/* with upstream sources, contrary to > what --merge-mode=replace should do.
That would be a grave bug. Can you still reproduce it? If so please send me the refs of the branches (master and upstream) before you run uscn so I can try to reproduce. > Sadly I wanted to keep working on guitarix and submit a new upload, so I > already committed 0.36.0 to the guitarix repo at > https://anonscm.debian.org/cgit/pkg-multimedia/guitarix.git/ . > > > Can you reproduce this with other packages? > > I tried to reproduce it with git-buildpackage 0.9.1 against python-pyo and > lv2proc packages, and it worked fine. > > So if you see no problem on gbp output as said, feel free to close the bug. > Maybe it was a fluke caused by several people working in guitarix's > repo. See above. Can you try to pass me the commits (or even better prepare a repo in the exact state) that I need to run gbp import-orig against? I tried several variants here but it always worked (as it did with the test run on the rest of the archive on my last sweep). Even the debian/ tree from your logs (c4a8a211261fc53b556732b1b724f938060d0135) is the same one that my invocation uses as is the parent commit on upstream. If you could provide more information on how to reproduce this that'd be great. If not let's close this for the moment. Should you hit that again please tar the _whole_ git repo and send me link so I can infer things from the reflog. Cheers, -- Guido
