(cc'ing the debhelper maintaner to get an authoritative answer about whether a dh module can have a "-" in the name or not).
On Tue, 03 Dec 2013, Axel Beckert wrote: > Matthias Klose wrote: > > Seen now in many packages, calling > > > > dh --with autotools-dev > > > > instead of autotools_dev, making this a nop. > > O.o > > Then this is a bad design choice by whoever decided that there should > be an underscore at that place I wonder if this is not because of a *technical* reason? Some stuff heavily objects to "-" in names. Still, maybe it was named like that because mostly the dh stuff uses _ to separate name components... Anyway, I am not the author of the dh helper, I accepted it in autotools-dev because autotools-dev is the appropriate package to carry it. Therefore, someone with the relevant technical knowledge will have to assure me that there are no issues with "-" in the dh module name so that I can come up with a proper way to address this issue. > Always dashes between "autotools" and "dev", but never an underscore. > So I'm not surprised at all that people get that wrong. I am quite appalled that people pay so little attention to their build that they don't even notice it *never worked correctly* in the first place. > I'd rather say this should be fixed in autotools-dev, at least in the > long run -- probably with some backward compatibility, e.g. via some > symlink. A single-place (i.e. autotools-dev + lintian check) fix is attractive, given the number of packages affected. However, the "nice to the wrongdoers" fix (alias autotools-dev as an alternative dh module name for the autotools_dev dh module) is only acceptable if I can also backport it to stable (and oldstable if at all possible), otherwise it becomes a silent landmine for backports (given that the wrongdoers won't have a proper versioned build-dependency). I don't mind doing the stable updates at all, but it requires a pre-approval by the release team. If they don't like the idea, I think we'd have to change autotools-dev to FTBFS anything that uses the wrong naming (maybe after a mass bug-filling and a grace window of a few months, whatever). > > Please add a lintian check for that. > > Definitely, at least until this is fixed in autotools-dev. You're likely have to carry a lintian check for this for at least two or three stable releases. No matter what fix we deploy, packages will always have to either use the old name, or declare a versioned build-dependency to be completely correct and backports-friendly... Granted, depending on the fix, it will just be a lintian warning. > Anyway, thanks for making me aware of that, I seem to have tons of > packages affected by this future lintian warning. Argh. Indeed we need to address this. -- "One disk to rule them all, One disk to find them. One disk to bring them all and in the darkness grind them. In the Land of Redmond where the shadows lie." -- The Silicon Valley Tarot Henrique Holschuh -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected]
