On Thu, Oct 02, 2003 at 08:36:50AM -0400, Stephen Frost wrote: > * Steve Langasek (vorlon@netexpress.net) wrote: > > It's been noted several times that the end of the 0-day NMU period was > > accompanied by a marked reversal in the RC bug graph. I think it's time > > for a group debriefing of this experience. I was pleasantly surprised > > to have not heard of a single complaint about bad NMUs during this > > period, either personally in response to my own NMUs or on the lists. > Where have *you* been? Right here, busily NMUing packages with beeswax in my ears to block out the screams... > The (attempted) NMU of epplets was *terrible*. > a) The NMUer got the *copyright* information wrong and just made all of > epplets appear as if under the GPL when it's certainly not (and the > primary authors do *not* like the GPL at all). Prior to the NMU the > copyright information was correct (portions under a BSD-style > license, and one specific module under the GPL). > b) The NMUer *never* sent any patch to the BTS even though quite a few > changes were done which I had to track down (including the copyright > fuckup). > c) Only some of the bugs which the NMUer fixed (amazing that any > actually were) were marked as having been fixed in NMU. > d) The NMUer apparently had a total lack of understanding when it came > to autoconf/libtool since he pretty much arbitrairly added them as > Build-Depends when they didn't need to be. > And epplets isn't exactly a hard thing to package. > Thankfully that was the only package of mine that was NMU'd. Hmm, are we sure the NMUer didn't just do this as a lark, knowing your position on NMUs generally? ;) The above sounds like a very bad NMU. But I don't think that's due to any lack of emphasis on good NMUing practices in the 0-day announcement; several of the points you complain about above were specifically prohibited under both the usual rules and the 0-day rules (no patch in the BTS, which should *always* be sent before the package hits incoming; changes for things not in the BTS; gratuitously intrusive changes). Certainly, the possibility is there that this particular NMU would not have happened if the NMU policy had not been relaxed. But honestly, for as long as this BSP lasted (and as many NMUs were done during the period[1]), the casualty rate seems to be a lot better than in previous BSPs where 0-day NMUing was *not* allowed. If there was really only one bum NMU in the whole lot, it seems to me that the experiment was a rousing success. -- Steve Langasek postmodern programmer [1] Looking just at the .changes filenames for the period, there appear to have been 312 sourceful NMUs of non-native Debian packages.
Attachment:
pgpsbRvuRMwSq.pgp
Description: PGP signature