Bug#1054630: dgit - cant import llvm-toolchain-15.
Ian Jackson writes ("Re: Bug#1054630: dgit - cant import llvm-toolchain-15."):
> I'll work on some kind of fix.
I have a fix. There's some minor reorganisation of the dsc import
code, to make my fix apply to whatever is wrong with the historical
info in the package changelog. So I won't be backporting this right
away as a stable release update. I will consider doing that after
it's been in testing for a while.
You should find that you can install the 11.4 .deb, which I intend
to upload later today, directly on whatever system you are using.
Or, build your own from the source; again, after I've uploaded.
Anyway, thanks for the report.
Regards,
Ian.
--
Ian JacksonThese 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.
Bug#1054630: dgit - cant import llvm-toolchain-15.
Peter Green writes ("Bug#1054630: dgit - cant import llvm-toolchain-15."):
> Correct me if I'm wrong here, but doesn't the import process
> only use information from the topmost changelog entry.
I'm investigsting now. The faulty changelog entry indeed isn't the
topmost one. However, dgit needs to use the changelog entry
corresponding to the (first) import of that upstream version, for its
synthetic commits representing the .origs.
(dgit tries to make these stable, as that improves the synthetic git
commit structure.)
I think the best thing is to use dummy data for the authorship, and
the date of the last non-broken changelog entry before the one it
wants. But perhaps that's too complicated and dgit should just use
today's data and dummy authorship if any of this archaeology fails.
I'll work on some kind of fix.
Ian.
--
Ian JacksonThese 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.
Bug#1054630: dgit - cant import llvm-toolchain-15.
Peter Green writes ("Bug#1054630: dgit - cant import llvm-toolchain-15."):
> Correct me if I'm wrong here, but doesn't the import process
> only use information from the topmost changelog entry.
I believe so. I'd have to check the code. I think the top changelog
entry information might be used for the "record in suite" pseudomerge.
Ian.
--
Ian JacksonThese 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.
Bug#1054630: dgit - cant import llvm-toolchain-15.
On 27/10/2023 11:37, Ian Jackson wrote:
Peter Green writes ("Bug#1054630: dgit - cant import llvm-toolchain-15."):
Now on the one hand, the changelog in llvm-toolchain-15 is indeed broken.
I filed a bug about that.
Thanks. (FMR that's #1051961)
On the other hand, the package was accepted by the Debian archive and
related tooling. Not being able to import it is blocking things up
for raspbian. I'd appreciate a way to downgrade this error to a
warning.
*sigh* what a mess. Quite so. I might be able to look at this in the
next few days.
It's not 100% straightfrward, because dgit uses that information to
construct the commit messages for the dsc import. So I think some
dummy information will have to be provided. Or maybe we should be
using the dsc Maintainer instead. Presumably the archive is better at
rejecting a missing dsc Maintainer.
Correct me if I'm wrong here, but doesn't the import process
only use information from the topmost changelog entry.
Bug#1054630: dgit - cant import llvm-toolchain-15.
Peter Green writes ("Bug#1054630: dgit - cant import llvm-toolchain-15."):
> Now on the one hand, the changelog in llvm-toolchain-15 is indeed broken.
> I filed a bug about that.
Thanks. (FMR that's #1051961)
> On the other hand, the package was accepted by the Debian archive and
> related tooling. Not being able to import it is blocking things up
> for raspbian. I'd appreciate a way to downgrade this error to a
> warning.
*sigh* what a mess. Quite so. I might be able to look at this in the
next few days.
It's not 100% straightfrward, because dgit uses that information to
construct the commit messages for the dsc import. So I think some
dummy information will have to be provided. Or maybe we should be
using the dsc Maintainer instead. Presumably the archive is better at
rejecting a missing dsc Maintainer.
Ian.
--
Ian JacksonThese 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.
Bug#1054630: dgit - cant import llvm-toolchain-15.
Package: dgit Version: 11.3 dgit can't import the current version of llvm-toolchain-15 dgit import-dsc ../llvm-toolchain-15_15.0.7-10.dsc ..workingbranch Dgit metadata in .dsc: NO git hash using existing llvm-toolchain-15_15.0.7.orig.tar.xz using existing llvm-toolchain-15_15.0.7-10.debian.tar.xz dpkg-source: info: extracting llvm-toolchain-15 in llvm-toolchain-15-15.0.7 dpkg-source: info: unpacking llvm-toolchain-15_15.0.7.orig.tar.xz dpkg-source: info: unpacking llvm-toolchain-15_15.0.7-10.debian.tar.xz dpkg-source: warning: unexpected end of diff 'llvm-toolchain-15-15.0.7/debian/patches/gcc-13-build-fix.patch' dpkg-parsechangelog: warning: debian/changelog(l124): found start of entry where expected more change data or trailer LINE: llvm-toolchain-14 (1:14.0.6-16) unstable; urgency=medium dgit: error: missing field Maintainer in package changelog, entry no.11 Now on the one hand, the changelog in llvm-toolchain-15 is indeed broken. I filed a bug about that. On the other hand, the package was accepted by the Debian archive and related tooling. Not being able to import it is blocking things up for raspbian. I'd appreciate a way to downgrade this error to a warning.
