Package: dgit
Version: 13.12
Simon Josefsson writes ("Re: [tag2upload 578] failed, python-securesystemslib
1.3.0-1"):
> Ian Jackson <[email protected]> writes:
> > No. I mean, it should have spotted the problem and failed a check.
>
> I did another upload of this package, and I ran into the same problem.
Thanks for writing in. So I saw.
I tried git-debpush from dgit.git#main on your package and it didn't
fail any check. dgit --gbp build-source reproduces the error message.
It's a long time ago, but my memory of the reason I wrote this in dgit
is as follows. I saw the frequent presence of packages with changes
to the toplevel .gitignore, but without any patch for that.
I inferred, perhaps wrongly, that gbp-based workflows are *supposed*
to omit such patches.
I experimentally removed this error check from my local dgit. When I
did that it was able to build a .dsc (and it passed the new git==dsc
correspondence check).
I propose that rather than changing git-debpush to detect this
situation, we change dgit (and thus the tga2upload service) to accept
it.
Then --quilt=unapplied will be (just) for people who want to *insist*
that their maintainer git tree *does* contain such patches, and
--quilt=gbp will tolerate it either way.
Sean, Simon, what do you think?
This change also applies to git-dpm. I propose to retain it there, at
least until we get a report from a git-dpm user. git-dpm has a
different model - generally, patches-applied - so it is probably
responsible for making (or not making) such patches.
The root cause of all this nonsense is #908747 in dpkg-source.
Ian.
--
Ian Jackson <[email protected]> These opinions are my own.
Pronouns: they/he. If I emailed you from @fyvzl.net or @evade.org.uk,
that is a private address which bypasses my fierce spamfilter.
