[fixing bug's address in Cc] Dmitry Bogatov wrote on Mon, 10 Dec 2018 19:49 +0000: > I believe, most of us keep repositories of git packages somewhere under > ~/. For example, for me, ucspi-tcp package is located at > /home/iu/devel/salsa/ucspi-tcp. > > So my workflow is following: > > $ cd /home/iu/devel/salsa/ucspi-tcp > $ dpkg-buildpackage -us -uc > $ lintian > > And here lintian could check, that generated binary packages does not > contains references to /home/iu/devel/salsa/ucspi-tcp.
So what's the rule? Lintian should check that the current working directory lintian runs from doesn't appear in the build output? This will false positive if lintian is run from the root directory, and also going to make lintian's own output dependent on the phase of the moon, in that . % lintian ../build-outputs/foo.deb . and . % cd ../build-outputs % lintian foo.deb . would produce different outputs. I would say that's undesirable. (This isn't a made-up example; 'pdebuild -- --buildresult=../build-outputs/' — that's a verbatim dot-dot meaning the parent directory, not an ellipsis — is part of my regular workflow.) However, the buildinfo file already includes the build directory in the Build-Path header. Would it be sensible for Lintian to check that the value of the buildinfo "Build-Path" header doesn't appear in the .deb's? Cheers, Daniel