Hi Aurelien, On Wed, Jan 08, 2025 at 07:57:42PM +0100, Aurelien Jarno wrote: > I can confirm that. That said I am a bit surprised that you only mention > these two. For instance libc6-dev-s390 and libc6-dev-i386 also both > conflict with libc6-dev-mipsn32.
I had problems importing mips64el as bookworm's qemu was not properly emulating it (bookworm-backports does). Hence my dataset missed mips64el at the time of reporting. > I guess there are many more conflicts on ports architectures, but I > guess your QA tools only look at the main archive. I have not looked into ports at all. > Yes, it's something possible, however given the above, I wonder if we > should just map the libc6-$arch conflicts to the libc6-dev-$arch ones. > It's sometimes overkill when there are no real file conflicts, but > that's probably easier to not miss any conflict and that doesn't break > anything either. Anyway, as you pointed above they are already not > co-installable so it doesn't break anything. I concur with mapping the Conflicts from libc6-$arch to libc6-dev-$arch. The negative effects of excessive Conflicts are hardly noticeable and the mirroring significantly eases maintenance down the road. Would you do the implementation or rather have me work out a patch? Helmut
