Re: [RFC] Restricting dependacies to particular platforms
Decklin Foster wrote:
> Martijn van Oosterhout writes:
> > I think we should find a way of marking these so all the
> > spurious warnings can be removed. Maybe:
> > Depends: slink:libasound | potato:libasound0.4
> > Depends: alpha:libc6.1 | !alpha:libc6
> I'm inclined to think that the program which checks for bad
> dependencies should keep track of this information itself. As you have
> the package lists for (slink, alpha, etc) already, it shouldn't be
> that hard. Right?
Well, if you had all the Packages files then you could easily
make sure that all packages that are mentioned exist somewhere
but this is not reliable for finding bad links.
Say if in the above example, both a libasound0 and a libsound0.4
exist for alpha, where is the bug? A missing libasound0 in potato
or is the libasound0 in alpha wrong.
If a libc6.1 is created for i386, does this satisfy the second
dependancy? Currently it does, but who know if it really works?
With the number of packages that changed names this is bad enough
tracking for just one architechture.
Anyway, I gather from this (non-)discussion that people think it
is a bad idea.