Re: [gentoo-dev] AMD64 Arch Testers needed urgently

[Kent Fredric]

On Thu, 14 Dec 2017 16:04:17 -0600
R0b0t1 <r030t1@×××××.com> wrote:

> Can you be specific? Human memory is biased towards negative
> experiences. If it's hard to actually describe the multitude of issues
> that mixed systems cause then it is very likely mixed systems do not
> cause many issues.

We have mixed system issues in part *BECAUSE* we use keywords
to migrate through transitions.

So the expectation is that when X gets keyworded stable, that diligence
has been performed to ensure everything else that needs to be stable at
the same time in order to retain function, is also stabilized.

Very much the case with Perl Virtuals.

The design of Perl virtuals is pretty much a bodge for Portage not having
a usable "provides" model and needing to represent certain upstream dynamics
in ways portage can't otherwise handle. ( Due to portage being "package" based
 and upstream being "component" based )


And for that to work, you pretty much need to keyword virtuals, and perl, 
and sometimes perl-core/, in flying formation.

And if you don't get those exposed to you in flying formation, portage gets
very confused and tries to do silly things.

( In my testing rigs I just don't bother with virtuals and have 
a >50 line package.provided file )

But that's just one example of how we use keywords in flying formation as
a mechanism for keeping parts working.

These all stem ultimately from the knowledge that other approaches
( eg: dependdency specifications ) all end in portage getting confused in
 dependency graphs and invoking impossible stabilization requirements.

Hence, keyword mixing is "discouraged if you want a stable working system".

If you're advanced and responsible however, go ahead, please do, and test.