Re: Debian GNU [was: smarter way to differ architectures needed?]
>>>>> Timothy Baldwin writes:
TB> I suggest that hwarch-foo can only be installed on hardware foo
TB> and provides hwiset-bar for each instruction (sub-)set foo
TB> provides.
It would be easier to call `hwiset-*' simply `cpu-*'.
In fact, that is the conclusion that Marcus Brinkmann led me to, and
so it will be in my policy proposal.
A lot of the complexity you are talking about is premature. Let's not
worry about this until we have some real-world examples, and not just
`foo' and `bar' to talk about.
I think that my proposal is sufficiently general so that it won't
block progress on these kinds of issues, which is the most that we can
ask for without going into a lot of detail.
TB> Currently we have libraries which share the same soname, but are
TB> binary incompatible (as they have been compiled for different
TB> architectures), this could be solved by putting a identifier for
TB> the binary interface in the soname or put them in different
TB> directories. It is also possible that in the future ld.so takes
TB> care of instruction set dependencies. User-Agent:
TB> POPstar/2.01b11
You have hit the nail on the head. This is the eventual goal, and
will be the solution to the huge libtool flamewar that we had on
debian-devel a while ago.
It will also make things like the libc5->libc6 upgrade a heckuva lot
easier.
TB> This solution solves most problems of the solution currently
TB> being discussed, except that does not cater for SMP systems
TB> [with] dis-similar processors.
That's a *slightly* harder problem than I seek to solve with my
initial proposal. ;)
>> >I think that an essential required base package should Provide
>> >default hwarch. Maybe that package should be `base-files'?
>>
>> I think base-files currently has architecture set to "all"...
>>
>> Anyway, I personally would prefer to keep them seperate.
TB> Especially as they could be very hardware dependant.
Exactly. And so, I'm in the process of designing an experimental
`hwarch' package which encapsulates all the Provides that the package
system needs to know about the physical machine.
Thanks for your comments: they definitely explore my ideas to their
logical conclusion.
--
Gordon Matzigkeit <gord@fig.org> //\ I'm a FIG (http://www.fig.org/)
Committed to freedom and diversity \// I use GNU (http://www.gnu.org/)
Reply to: