Grar, this was a lot more confusing than I expected (partly because the file that was uploaded to Launchpad for the 2012.04 release is _not_ the same as the file on pypi).
What seems to happen is this:
1. lp:lava-android-test does not contain a setup.cfg file
2. when you run python setup.py sdist, this runs egg_info and so creates a SOURCES.txt that does not contain setup.cfg
3. about the final thing sdist does before making the tarball is write a setup.cfg (this is a setuptools thing)
4. when you unpack the tarball, debianize it and run dpkg-buildpackage, this runs egg_info, which now updates SOURCES.txt to contain setup.cfg
5. when you run dpkg-buildpackage again dpkg-source complains because the tree has been modified.
I guess this is a setuptools bug really. But the workaround is easy: commit a setup.cfg file to lp:lava-android-test (most of our projects have one anyway to turn gpg signing on)
Grar, this was a lot more confusing than I expected (partly because the file that was uploaded to Launchpad for the 2012.04 release is _not_ the same as the file on pypi).
What seems to happen is this:
1. lp:lava-android-test does not contain a setup.cfg file
2. when you run python setup.py sdist, this runs egg_info and so creates a SOURCES.txt that does not contain setup.cfg
3. about the final thing sdist does before making the tarball is write a setup.cfg (this is a setuptools thing)
4. when you unpack the tarball, debianize it and run dpkg-buildpackage, this runs egg_info, which now updates SOURCES.txt to contain setup.cfg
5. when you run dpkg-buildpackage again dpkg-source complains because the tree has been modified.
I guess this is a setuptools bug really. But the workaround is easy: commit a setup.cfg file to lp:lava-android-test (most of our projects have one anyway to turn gpg signing on)