Sean Whitton writes ("Bug#848931: source-only builds and .buildinfo"):
> I think that dgit should strip the .buildinfo when it built the source
> package itself, as part of a build-source or push-source run, but what
> about when the user supplies a .changes, with -C, to push-source?
>
> I don't think that dgit should modify the .changes in that case, because
> it is counterintuitive when the user has provided a particular .changes
> they want dgit to verify and then upload.
Yes.
> Maybe it should refuse to
> push if the .buildinfo is present, though? Possibly with a
> --deliberately-include-buildinfo to override.
dgit should IMO refuse if there is an <arch> buildinfo which would
prevent a buildd upload (see recent debian-devel discussion).
(I think this may need to be configurable at some point so we should
try to avoid making that too difficult.)
> W.r.t. building source packages, do you still think that we should be
> using dpkg-source to build it, or just continue invoking
> `dpkg-buildpackage -S -us`? This is relevant for both build-source and
> push-source.
I think using dpkg-source is probably better. We have to cope with
changes to the behaviour of dpkg-source already. Whereas, we want to
avoid being broken by unexpected changes in dpkg-buildpackage.
Ian.
--
Ian Jackson <[email protected]> These opinions are my own.
If I emailed you from an address @fyvzl.net or @evade.org.uk, that is
a private address which bypasses my fierce spamfilter.
