Hello, On penktadienis 29 Liepa 2011 15:24:10 Raphael Hertzog wrote: > On Mon, 25 Jul 2011, Modestas Vainius wrote: > > What do you think about taking this further and supporting expressions > > like cpuattr:value (or cpuattr-value, or cpuattr=value, or whatever > > other syntax) in the architecture wildcards (Dpkg::Arch::debarch_is()). > > In this particular case: > > > > * endian:little and endian:big > > * bits:32 and bits:64 > > > > As a result, arch symbol tag, [] expressions and Architecture fields in > > debian/control etc. would gain support for this automatically. In my > > opinion, this would be more flexible in the long term. > > The idea seems nice but I'm not sure it's going to make the implementation > any cleaner.
Yeah, probably implementation won't be cleaner (but neither complex) however, in my opinion, this would improve flexibility of the expression itself. For example, size_t expands to 'long' on all 64-bit arches and s390. This could expressed as: (arch=cpubits-64 s390) Likewise, it would be possible to match an arch in the negation manner. E.g. all 64bit arches on linux: (arch=!cpubits-32 linux-any !any) I must admit, the latter isn't a very trivial and obvious expression but still a possibility. Could we improve it? Now if this was as: (arch=s390|bits=64) it might look a bit confusing. Sure, "|" here is a tag separator but most will assume it to be a disjunction so it would a be mistake to imply anything else. This would leave us no way to implement negation. > And I don't see really the need for it in other places where > we currently use architecture wildcards. I agree. I couldn't think of any real world example either. I just want an expression to be powerful enough as weird arch-specific symbols are not a very rare case. -- Modestas Vainius <mo...@debian.org>
signature.asc
Description: This is a digitally signed message part.
