[debhelper-devel] Bug#766795: afterstep not binnmu safe and not installable in sid

Simon McVittie smcv at debian.org
Sun Nov 2 13:46:02 UTC 2014


On Sun, 26 Oct 2014 at 00:36:29 +0200, Robert Luberda wrote:
> > It seems after the binnmu for the libjpeg-turbo transition afterstep is
> > no longer installable in sid (and hence the binnmu won't migrate and the
> > transition won't complete in testing).
> 
> This is debhelper bug. I've just uploaded afterstep 2.2.12-3, which
> removes usage of {misc:Depends} as a work-around.

I'm not so sure that this isn't a bug in afterstep, actually.

Summary of the bad situation: suppose we have an arch-independent
package, foo-common, and an arch-dependent package, foo-bin. If foo-bin
uses dh_installdocs --link-doc=foo-common then bad things occur.

The current bad thing is that foo-bin Depends: foo-common (= ${binary:Version})
which is not binNMU-safe: when foo-bin is binNMU'd, foo-common will not
satisfy the dependency.

The solution proposed by Robert Luberda seems to be to change debhelper
so that foo-bin Depends: foo-common (= ${source:Version}) instead,
in the case where foo-common is arch-independent. However, that would
mean that the changelog.Debian.amd64.gz (or whatever) for foo-bin
is not provided by anything, so the binNMU is not documented anywhere.
Is that itself considered to be a bad thing, and possibly a policy violation?

The solution I used in telepathy-glib was to link the documentation of
all arch-dependent packages to the shared library (which in practice
they all depend on anyway), and leave the one arch-independent
package alone:

	override_dh_installdocs:
		dh_installdocs -a --link-doc=libtelepathy-glib0
		dh_installdocs -i

Is that considered to be best-practice?

    S



More information about the debhelper-devel mailing list