Sean Whitton writes ("Bug#886969: Bug#886625: push-source should be usable no
matter the state of the working tree"):
> On Sun, Apr 22 2018, Sean Whitton wrote:
> > I thought of a case where a .dsc is not sufficient and one needs a
> > .changes: services like <http://debomatic-amd64.debian.net/>.
>
> I thought of one more case recently: uploading to security-master. You
> can't dgit push there yet, so you need an _source.changes.
You can't do a source-only upload either, so you have to do something
like dgit sbuild. (Obvs. you don't want to upload to -security
binaries made in your random workstation environment.)
I have been using dgit sbuild -A, and then dput followed by a push to
salsa. It's quite unsatisfactory really (but better than the
alternatives).
> This might be a reason to not remove but instead firmly deprecate
> build-source, until one can dgit push to security-master.
I don't really mind deprecating it. Dropping it entirely would be a
different matter. It might be embedded in people's workflows. FTR
I'm the kind of person who takes ages to remove old stuff.
I agree that there should be dgit export-dsc, and that it should be
possible to dgit push-source with a dirty tree by using sufficiently
vigorous command line options.
> The reason for limiting dgit subcommands in this way is to
> restrict dgit's role to being the bidirectional archive<>git
> gateway, not an all purpose Debian packaging wrapper script like
> git-buildpackage is.
I keep saying this: I would love for dgit not to be a general purpose
wrapper but until the .gitignore bug is fixed in all other tools it
will have to continue to be so (and then for some time afterwards).
(Also, -wgf.)
I mind much less that dgit is a general purpose wrapper (even though I
have to maintain and test a stupid pile of wrapper and command line
parsing code) than the fact that users can't use their existing build
tooling.
Also in this area is the need for a way to do an sbuild on a git tree
which is not a fixed point under (or, even, representable by)
dpkg-source. There's a bug about that too.
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.
