[git-buildpackage] What is the optimal workflow for git-buildpackage with dfsg-repacking
Guido Günther
agx at sigxcpu.org
Wed Dec 11 09:04:24 CET 2024
Hi,
On Tue, Dec 10, 2024 at 09:57:57PM -0800, Otto Kekäläinen wrote:>
>
> We discussed this a little over a week ago:
>
> > > > > > > > Ref:
> > > > > > > > https://salsa.debian.org/otto/box64/-/blob/debian/latest/debian/gbp.conf
> > > > > > >
> > > > > > > I'm not using `ignore-branch` to prevent errors like uploading from the
> > > > > > > wrong branch.
> > > > > >
> > > > > > How does your contributors send you Merge Requests? What branch names
> > > > > > do they use?
> > > > >
> > > > > Usually they use a topic branch name. It doesn't usually matter though.
> > > >
> > > > What packages are those about? Are they running Salsa CI? If you have
> > >
> > > salsa-ci yes.
> > >
> > > > a hard-coded branch in gbp.conf, the Salsa CI would always run on the
> > > > debian-branch and never actually test the change the submitter did,
> > > > right?
> > >
> > > I don't understand what you mean by "hardcoded branch" in gbp.conf and
> > > how that would affect salsa-ci.
> >
> > So locally obviously all contributors would have to start passing
> > manually --ignore-branch to get past the initial gbp error:
> >
> > ± gbp buildpackage
> > gbp:error: You are not on branch 'debian/latest' but on 'test-ignore-branch'
> > gbp:error: Use --git-ignore-branch to ignore or --git-debian-branch to
> > set the branch name.
>
> ..and as a result I stopped using 'ignore-branch = True' in new
> packages as I thought it is not needed anymore. However today I came
> across a contributor that ended up doing a completely new MR just to
> rename the branch they were using to `debian/latest` to be able to
> build it: https://salsa.debian.org/debian/godot/-/merge_requests/6#note_558493
>
> So I switched back my opinion to this: 'ignore-branch = True' should
> be default to make it easier for contributors to send Merge Requests.
> Forcing them to discover and pass manual workarounds on every build
> invocation is unnecessary friction on a path that is anyway full of
> all kinds of Debian quirks. The "protection" is provides for the
> occasional source-only upload-build is much more rare than the amount
> of friction it creates for 99% of normal builds while working on
> packaging.
I can't stop you from doing but want to note that the option to pass is
printed *in* the error message.
Cheers,
-- Guido
More information about the git-buildpackage
mailing list