Proposing new dual booting release criteria

[Adam Williamson]

On Mon, 2014-09-29 at 10:16 -0400, Gene Czarcinski wrote:

> > 2. "The installer must be able to install into free space alongside an
> > existing OS X installation, install and configure a bootloader that will
> > boot Fedora; if the boot menu presents OS X entries, they should boot OS
> > X."
> >
> > (so far as I could see on a quick skim back through the thread, this was
> > the most recent version of the OS X proposal). I am +1 to this too, it
> > seems reasonable. We could perhaps insert that the Fedora install
> > process should not render the OS X install unbootable from the EFI boot
> > manager?

> Now I am all +1 for being able to multi-boot Fedora an OS X on the same 
> hardware but I though the Chris Murphy (who is somewhat of an Apple User 
> expert) said (somewhere/sometime) and anaconda/Fedora was getting it all 
> wrong and not reliable at all *and* that a better way was ("it hurts so 
> don't do that") to not use the Fedora's bootload but instead use the 
> firware bootloader to boot OS X.  Just saying.  Now if grub2 can be 
> fixed so that it will reliably boot OS X, then +1.

This is the version that actually covers that - note the "if the boot
menu presents OS X entries". The "if" implies that it's fine for the
boot menu *not* to present OS X entries and rely on the EFI boot manager
for access to OS X. My general rule of thumb is that someone mistaking
the intent of something like a criterion means it's not clear enough, so
- perhaps we could explain this as a footnote?

> > 3. The installer must be able to install into free space alongside an
> > existing GNU/Linux installation, install and configure a bootloader that
> > will boot both systems, within the limitations of the upstream
> > bootloader."
> > Within the limitations? [show] Purpose of this clause is to not require
> > us to fix upstream bootloader bugs or design limitations.
> >
> > This is the complex one we're still struggling with. I think the above
> > is possibly a little broad and could do with either limiting to
> > stock-ish installs of 'commonly-used' or 'popular' distributions, or
> > some more vaguely-worded wiggle room clause. I don't want to have to
> > come up with some kind of criterion judo to justify us not slipping
> > Final release three weeks to fix, I don't know, dual-boot with an xfs
> > install of Fermi or something (no disrespect intended, Fermi users...)
> While supporting other distributions would be nice, I would consider 
> that (as a minimum) I should be able to install Fedora along side other 
> installations of Fedora.
> 
> Q: What about installations for Fedora which use previously (Fedora or 
> at least Linux) allocated partitions/LVMlv/btrfs-subvolumes?

That's the kind of thing that's very grey area-ish. I suspect we're not
going to be able to formulate this completely precisely, it's going to
be one of those ones where we have to have some subjectivity. I'm not
sure we want to get into prescribing partitioning behaviour.

> Q: As currently envisioned, what magic is going to be used to boot these 
> other systems?  Currently, grub2's /etc/grub.d/30_os-prober is broken at 
> least as far as booting other Fedora 21 installations:
> https://bugzilla.redhat.com/show_bug.cgi?id=1108296

The criteria aren't about the implementation, they're about the design /
requirements. Generally speaking, it's usually a good idea to assume
inertia: if no-one explicitly causes things to change, Fedora will
continue using 30_os-prober, that's just the way of the world. That
would mean it would have to get fixed with this change.

cmurf, you ran this proposal by the anaconda list, didn't you? How did
that turn out?
-- 
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | XMPP: adamw AT happyassassin . net
http://www.happyassassin.net