Hi, On Fri, Jan 10, 2020 at 04:58:22PM +0100, Christoph Biedl wrote: > This leaves me somewhat confused since I understand your rationale > the file package needs itself to be built, in other words, a circular > build dependency. This should not happen, and I'd like to eliminate > that. > > Mind to shed some light on this? How was libmagic wrongly built, how did > this manifest?
Well, the circle is implicit. file is used by debhelper/dpkg to generate shared library dependencies. libmagic1 misses its library dependencies on a number of architectures. Aurelien suggested that someone greps .buildinfo files on coccia for packages built with 1:5.38-2: ./01/06/girara_0.3.4-1_all.buildinfo ./01/06/girara_0.3.4-1_amd64.buildinfo ./01/06/girara_0.3.4-1_arm64.buildinfo ./01/06/girara_0.3.4-1_armel.buildinfo ./01/06/file_5.38-3_amd64.buildinfo ./01/06/file_5.38-3_arm64.buildinfo ./01/06/girara_0.3.4-1_armhf.buildinfo ./01/06/girara_0.3.4-1_i386.buildinfo ./01/06/file_5.38-3_armel.buildinfo ./01/06/girara_0.3.4-1_ppc64el.buildinfo ./01/06/girara_0.3.4-1_s390x.buildinfo ./01/06/file_5.38-3_armhf.buildinfo ./01/06/file_5.38-3_i386.buildinfo ./01/06/file_5.38-3_mips64el.buildinfo ./01/06/file_5.38-3_ppc64el.buildinfo ./01/06/file_5.38-3_s390x.buildinfo ./01/06/python3.8_3.8.1-2_amd64.buildinfo Looks like we want to binNMU girara and python3.8 as well. Helmut