On 10/03/2011 12:37 AM, Chí-Thanh Christopher Nguyễn wrote: > Samuli Suominen schrieb: > >>> And again, downgrade of dependencies it is not against any rule which >>> would justify mask and removal. >>> >>> Another example from the X.org packages, installing the proprietary >>> ATI/NVidia drivers will cause downgrades for xorg-server on ~arch >>> systems. Nobody in his right mind is proposing to treeclean them because >>> of this. >>> >> >> The new xorg-servers could get package.masked until these major drivers >> are available. >> Albeit, I'm not intrested in pursuing this since with separate >> xorg-server package, it's the drivers that need rebuilding against it, >> and the VIDEO_CARDS="" setting is keeping it in certain version until >> the VIDEO_CARDS="" setting is satisfied. >> >> Poor example to make a case. > > VIDEO_CARDS is just for user convenience. run "emerge nvidia-drivers" on > any system with xorg-server-1.11 installed and it will downgrade, no > matter what VIDEO_CARDS is set to.
And your point is? The drivers will need to be rebuilt everytime the xorg-server version changes. This does not come as a suprise, the .ebuild should print a message about rebuilding them. If it doesn't, then the .ebuild should get fixed. Leaving this particular case for X.org maintainers to decide sounds fine to me, given the relaxing factors. > >> The intresting part of that document is "You should also not cause an >> unnecessary downgrade for any "~arch" when ..." which also applies to >> setting dependencies just as well. > > The downgrade is necessary to avoid user-visible breakage. Avoiding one in non-system critical package (like qutecom), but introducing multiple new scenarios in what-could-be system-critical packages. > And the wording clearly does only apply to package removals. The fact that the *common sense* snippet was inserted in this document, but isn't documented else where... doesn't make it any less true. - Samuli
