[PATCH 1 of 2 logilab-mtconverter] Update Debian packaging
jeremy.bobbio at irq7.fr
Wed Jun 5 20:55:47 CEST 2019
On 05/06/2019 20:21, Denis Laxalde wrote:
> Jérémy Bobbio a écrit :
>> # HG changeset patch
>> # User Jérémy Bobbio <jeremy.bobbio at irq7.fr>
>> # Date 1559736762 -7200
>> # Wed Jun 05 14:12:42 2019 +0200
>> # Node ID fad960015fd7e9108483c23b590c03326efbce07
>> # Parent 9ad8571111d41592bd028eb52791cc601813e79b
>> Update Debian packaging
> Series applied, thanks.
>> diff -r 9ad8571111d4 -r fad960015fd7 debian/source/format
>> --- /dev/null Thu Jan 01 00:00:00 1970 +0000
>> +++ b/debian/source/format Wed Jun 05 14:12:42 2019 +0200
>> @@ -0,0 +1,1 @@
>> +3.0 (quilt)
> One question: can you explain this hunk? (This applies to all your
> patches so far, not just this one.)
> IIUC, this makes the package non-native. However, I'm not a specialist
> but I wonder if the native option wasn't actually intentional (since the
> vcs repo contains the debian/ directory, I guess). What's your (and
> others') opinion on this?
These pieces of software are available outside Debian (PyPI at least)
and works without Debian. The version in debian/changelog actually
reflects this: it specifies the upstream version and a package
revision (separated by an hyphen). In case a fix is made to
`debian/control`, there should be no need to require a new upload to
The canonical way to distribute a source tarball for a Python package is
by running `python3 setup.py sdist`. This makes a nice orig tarball to
build a Debian package against.
The last bit that might worth explaining then is switching from “1.0” to
“3.0 (quilt)”. There are two reasons to prefer the latter:
The `debian/` directory is shipped in a tarball instead of a diff. This
allows to put binary files, e.g. icons or data for a testsuite without
having to use base64 or another text encoding.
Meanwhile, it requires changes to the upstream source to be properly
recorded. This makes it harder to have erroneous changes go unnoticed.
I hope that makes enough sense. :)
More information about the cubicweb-devel