On 17-02-13 20:35, Paul Gevers wrote: > On 17-02-13 19:16, Graham Inggs wrote: >> Ah, I didn't notice that the transitional libmotif4 provided libmotif3. >> I have removed that, because it is a transitional package it can be >> uninstalled after the transition, but we still need to show that our >> libxm4 (and libmrm4 and libuil4 and libmotif-common together) still >> provide libmotif3 because of the lib*.so.3 symlinks. >> >> Regarding the moving of the Provides from the Multi-Arch: foreign >> libmotif-common to the Multi-Arch same libxm4, I thought of a scenario >> where some could have, for example, only the x86_64 version of libmotif4 >> installed, they then upgrade to our new package which now provides >> libmotif4 in a Multi-Arch: foreign package. >> They then install an old package depending on libmotif:i386. The >> installation would proceed without pulling in the i386 libxm4. etc. > > My proposal is different: > libmotif4: provides libmotif3 (I added in the description that it can > only be removed if no packages depend on libmotif3) and is indeed > multi-arch same. > > All other packages CAN NEVER provide libmotif3 as for each package it > will always be incomplete, and thus no other package can provide full > libmotif3 support on it's own. To be perfectly clear, I think ONLY libmotif4 could ever provide libmotif3. The alternative is a libmotif3 package. This is independent of our final choice of multi-arch for each package. Paul
Attachment:
signature.asc
Description: OpenPGP digital signature