Re: cameleon: byte or native?

[Sven Luther <luther@dpt-info.u-strasbg.fr>]

On Thu, Nov 28, 2002 at 08:46:49AM +0100, Jérôme Marant wrote:
> >The policy says it chooses one at random, the implementation says it
> >chooses the first by default.
> > 
> >
> 
> Strange. I recall that sometimes it tells the user to specify the wanted 
> package (among those alternatives) in the command-line. Am I wrong?

Not sure, i don't remember exactly, but the behavior i describe
(choosing the first one) is the one implemented by the autobuilders, and
is why i do :

Build-depends: xlibmesa-dev  libgl-dev (mmm, it is false though, i only
have bgl-dev)

for lablgl for example.

Also, apt-get will privilege a real package over a virtual one, and that
is why our native/byte code separation thingy will work. I have this
from one of the apt authors, but it would be nice to have it in policy.

I think the asking thingy you speak about is when you only have virtual
packages or something such, not sure though.

Now all this is fine, but does not really respond to your question, so i
suppose we will have to test it, but again,  for the autobuilders it
_will_ work.

If you want to test it, just build a package that depends on spamoracle,
(or add a spurious spamoracle | spamoracle-byte dependency to one of
your packages) and try either and apt-get build-dep and/or building it
with pbuilder to see what happens.

It would be nicer if we had something written in the policy about that,
maybe there even is something about it already, we would need to check
about this.

Friendly,

Sven Luther