Re: [gentoo-dev] Virtuals revisited (Round 3)

[Jason Stubbs]

On Thursday 07 April 2005 23:26, Ciaran McCreesh wrote:
> On Thu, 7 Apr 2005 22:20:43 +0900 Jason Stubbs <jstubbs@g.o>
> wrote:
> | Back in my quest to rid the tree of the evil virtuals and their plot
> | to  destroy my mind. I think I've covered all bases with this one -
> | perhaps even  enough to satisfy TGL. Take your best shot!
>
> 1) You need an explicit example of what virtual/jdk/jdk-1.4.ebuild looks
> like.

Will add.

> 2) For the consistency section, I'm not sure how you're going to
> implement this. The example I tend to use is this (versions simplified
> slightly for clarity):
>
> * User emerges vim-6.3, gvim-6.3. ~vim-core-6.3 is in as an RDEPEND.
> * User upgrades to vim-7.0. ~vim-7.0 is in as an RDEPEND.
> * User still has gvim-6.3 installed. gvim-6.3 has an RDEPEND upon
> ~vim-6.3, which is no longer installed. gvim is now broken.
>
> An example of exactly how portage would 'solve' this would be
> helpful. Would it ask for an unmerge of gvim-6.3 or would it include an
> upgrade of gvim-6.3 as part of the "emerge ~vim-7"?

Yep something like that. The specifics of what to do in various cases such as 
that one have not been worked out but are irrelevant to the glep anyway. The 
glep only need require that it is handled.

> 3) We don't specify every virtual in every profile thanks to cascades.

And...? I guess you are referring to "profiles must provide a default ..." but 
that _is_ a correct statement. If somebody wants to create a new profile that 
doesn't derive from base, they must specify defaults for all the virtuals.

> 4) Backwards compatibility -- all virtuals ebuilds should include
> DEPEND=">=portage-2.whatever" as well as the "|| ( foo bar )" stuff.

Why? The virtuals would be no different to regular packages.


Regards,
Jason Stubbs
--
gentoo-dev@g.o mailing list