On Sun, Nov 20, 2016 at 03:12:41AM +0100, Guillem Jover wrote: > > dpkg-deb: building package 'myspell-fr-gut' in > > '../myspell-fr-gut_1.0-31_all.deb'. > > dpkg-genbuildinfo --build=all > > dpkg-genbuildinfo: error: cannot fstat file > > ../ifrench-gut_1.0-31_amd64.deb: No such file or directory > > dpkg-buildpackage: error: dpkg-genbuildinfo --build=all gave error exit > > status 2 > > -------------------------------------------------------------------------------- > > > > The package does not use dpkg-genbuildinfo explicitly, so this seems > > like a bug in dpkg-dev. > > > > On the other hand, if this were a bug in dpkg-dev, I would expect to > > find a lot more packages with the same problem, but I only found this > > one. > > > > So this is really strange and I don't really know where is the bug. > > Please do agree on which package is to blame. > > This package has broken debhelper usage, it does not specify neither > -a nor -i flags for debhelper commands, so the DEBIAN/control file > gets generated for both all and any packages, and dpkg-gencontrol > registers the to be generated .deb into debian/files, but then those > never end up being generated.
However, this is the first time that the package fails and other tools like dh_md5sums do not seem to have a problem with this. Considering that dpkg-genbuildinfo is called with --build=all (as shown in the build log), is this not a little bit inconsistent? I mean: What is the purpose of passing --build=all to dpkg-genbuildinfo if dpkg-genbuildinfo does not use this information? While we are at it, even if this was a bug in the package, would not make sense to have a dh_buildinfo which calls dpkg-genbuildinfo and honors the usual internal debhelper variables that make things like this not to happen with other dh_* commands? (I agree that this is not a big issue because the number of packages broken by this is very low, I only found two other packages with a similar bug so far). Thanks.
