On Sunday 27 April 2008, Adam D. Barratt wrote: > On Fri, 2008-04-11 at 20:29 +0200, Frans Pop wrote: > > There can be packages for which one has commit access and thus is > > allowed to (and even supposed to) add or create changelog entries, but > > is not an Uploader. > > > > Example in my case is debian-cd, to which I frequently make changes > > that are related to the installer, but for which Steve McIntyre is the > > main maintainer and which I've never uploaded. > > > > Having dch make commits into NMUs in that situation is completely > > wrong. > > I agree that it would be preferable if simply running "dch" always dtrt. > > To that end, we've just made a change which means that running dch on an > UNRELEASED changelog will skip auto-NMU detection if you're using the > "changelog" release heuristic (which you'd need to be in order for dch > to automatically add new entries to an UNRELEASED stanza anyway). > > Hopefully that goes some way to alleviating your concerns.
Not really as there is absolutely no reason why I can't be the first person to commit a change after a release (and thus be the person who opens a new changelog). I've now seen so many exceptions and extra checks and whatnot being implemented for the auto-NMU change that that only strengthens my conviction that it should never have been made default in the first place. Sure, it's a useful option for those very few people who do frequent NMUs for example during BSPs, but it seems to me that for the fast majority of users the "-n" option is the indicated method to set an NMU version number instead of this "let's try to guess and sorry for the mess if we guess wrong" approach that's auto-NMU. IMO the "-n" option is a 100% usable, obvious and correct method to open a new changelog entry for an NMU without any concerns about debchange ever doing the wrong thing. I repeat my request: please disable auto-NMU by default and allow those few users who really want it to enable it through ~/.devscripts. Cheers, FJP
signature.asc
Description: This is a digitally signed message part.
