Gergely Nagy: > Control: tag -1 pending > > Hi! > > To summarize the case quickly for Niels: if we have a package that has > an arch:all and an arch:any binary, and uses dh_install --fail-missing, > and dh-exec in either (or both) binary packages' .install file, and we > do an arch-indep or arch-only build, --fail-missing will break. > > [...] > > I'm Cc-ing Niels too, in case he has a better idea, and to see if he > sees anything wrong with the approach I may have missed. > > I plan to upload dh-exec sometime next week (likely at the end of it, > too), but I'd prefer to hear some feedback from the debhelper side too. > > Thanks for reporting the bug, and for Niels in advance for his input! >
Hi, I am at the point, where I believe that "dh_install --fail-missing" (or --list-missing) is at the wrong place. Notably, it will complain about files that will be installed by other helpers later (e.g. dh_installman). This is wrong, hard to fix in dh_install and the most obvious why dh_install is the wrong place for this option. I will have a look at fixing the debhelper side of this, so it becomes less magic. As it is, I cannot quite grok the proposed solution in dh-exec - which I am pretty sure is no fault of dh-exec itself, but rather an artefact of the entire --fail-missing interface. I wish I could say this meant we would not need all of this special-casing between dh-exec and debhelper. In fact, it will probably become more apparent as more helpers will have to do something similar. But I hope to make things simpler for all parties. Thanks, ~Niels
