Hi Ian! > > a cowbuilder hook) can see lintian errors in the source. I missed a > > few lintians because of it some time back, and then I stopped using > > "gbp cowbuilder" as a consequence. > > Mmmm.
sorry, should be "dgit cowbuilder". > > In other words, the .changes file produced does not list the .dsc and > > .orig.tar.gz files. Even when adding the dpkg-buildpackage switch -F, dgit > > seems to override it. > > Yes, indeed it does. It builds the source separately, and merges that > with the builder-generated changes. > > > Do you think it would be possible to add a command line switch for this? Or > > maybe just respect whatever "dpkg-buildpackage switch" is chosen? > > The dpkg-buildpackage switches control the generation of source > packages, but not anything to do with lintian. So that isn't the > answer I think. So the lintians for the source package don't appear if manually running lintian /var/cache/pbuilder/result/mypackage*.changes on a package built with "dgit cowbuilder". But if you build with normal cowbuilder they appear. This problem in itself therefore has nothing to do with triggers. > > I think this applies for the other "builders" in dgit also (sbuild, > > pbuilder..). > > Yes. In the function "massage_dbp_args" in dgit, the argument is this: # Since we split the source build out so we can do strange things # to it, massage the arguments to dpkg-buildpackage so that the # main build doessn't build source (or add an argument to stop it # building source by default). I do not completely understand. What would fall apart if you just let dpkg-buildpakcage build the source? Cheers Ruben
