Hi Vincent, On 24-08-15 23:52, Vincent Blut wrote: >>>> Which file do you have in common with ntp? Please re-read ¹. >>> >>> I guess I’ve been misled by § 7.6.2! The previous section shows the >>> usage of the 'replaces:' field for packages providing *files* already >>> provided by another package. However, the section 7.6.2 seems to be >>> for *packages* that /do conflict/; I interpreted that /do conflict/ >>> by "packages providing the same functionality". I even was quite sure >>> my interpretation was correct after seeing the usage example about >>> MTAs. >> >> I didn't check if ntp is also doing the conflicts/replaces/provides >> dance on time-deamon. If so I than you don't need to mention ntp >> specifically at all¹¹. > > It does not. There is still an opened bug report² about adding ntp to > the time-daemon > virtual package, but the discussion has stalled since few years now. > >>> Anyway, depending on your answer, I’ll revert this commit. >> >> Lets not do that until we agree. :) > > Ok. Hmm, so I believe "Conflicts" or maybe even "Breaks" is indeed enough. Once ntp provides time-daemon, that can be removed as well. >> Oh, you not understanding me makes sense. It was me who didn't >> understand what you were doing. I was assuming there now was a new >> mechanism, but now you explained it is just a better implementation of >> "the same thing". But then again, maybe make that a little clearer in >> your changelog for others like me? > > Ok, maybe adding something like: > > “Basically, this directive makes the detection of the standard (Local or > UTC time) set in /etc/adjtime — and used by the hardware clock — clearer > compared to the text processing method we used to use in the post install > script to complete the same task.” > > What do you think? Sounds like some good text (but maybe a little long for the changelog. I believe the NEWS file is there for this purpose. >>>> Can you please explain me how commit 1ce86d3 works (the Breaks of >>>> util-linux). >>> >>> As the hwclockfile directive can only deal with /etc/adjtime, we need >>> to ensure that we migrated from /etc/default/rcS to /etc/adjtime. To >>> be honest, I’m not sure that break is even needed, because this >>> migration happened prior to Wheezy. >> >> I haven't looked it up, is util-linux in essential? Otherwise, shouldn't >> you depend on it with the higher than dependency? > > Indeed, util-linux is essential hence the fact it is not listed in the > 'Depends:' field. Thus I think "Depends: util-linux (>= 2.20.1-5) is more correct. It is in essential, but you require a version higher than possible (when was 2.20.1-5 introduced?). >> You confirm my ideas here. But as I mentioned in my other mail (below), >> I challenge you to come up with a way to run them which is acceptable in >> Debian's autopkgtest framework. > > Challenge accepted, but I’ll have to document myself about autopkgtest, > especially on the integration of upstream tests. Great. Not needed for this upload per se. So, please fix the dependencies in git (I don't need the dsc, just ping me when your done) and I will build and upload. Paul
Attachment:
signature.asc
Description: OpenPGP digital signature