Re: GPT partlabel support for Mach
наб, le lun. 15 juin 2020 18:55:43 +0200, a ecrit:
> On Mon, Jun 15, 2020 at 02:46:47PM +0200, Samuel Thibault wrote:
> > No, that's the contrary: the set of packages included in the netinst
> > image is a coherent view at some point. The daily images have not been
> > tested so they might break, but the released images (be the actual
> > 2-year-period release, or the snapshot I publish from times to time)
> > have been tested, so the base system included in it *is* installable.
> Point. I was mostly thinking WRT conflicts between what's on
> the ISO vs what's in the archive, like the glibc versioning problem from
> January's image,
Ah, right, that could bring problems indeed, but usually at the apt step
it would be able to just upgrade with sid and unreleased.
> > The mini.iso image, however, *completely* depends on the archive state
> > at the time you run it. There is thus zero guarantee whether it's
> > installable at all or not.
> FWIW, I do most (all) of my installs over PXE (=> no local repo) or by
> strapping and it's always worked for me on release arches,
Release archs are much more taken care of, and use dak, which keeps the
old arch:all packages.
> > So you here mean the initial base system installation? With netinst it
> > shouldn't actually be needed to enable unreleased, it should be already
> > finding everything it needs inside the netinst iso for installing the
> > base system.
> Exactly, that's the magic I was referring to. It finds packages from
> entirely unrelated suites for the file:// archive, but not a remote one;
> this alone is weird, but I understand why it would be desired.
The CD doesn't really have a "suite" label, it's just a pile of .debs,
which I indeed pull from both sid and unreleased, to get something
installable.
> One solution would be to always enable unreleased for hurd
> (or, going by the check in apt-setup/50mirror[1], all ports arches?
That should already be the case for all ports archs, yes.
> a glance at the archive[2] says most of them have *something* there)
> in the debootstrap script, similar to what I sent before.
For the debootstrap call we'd have to fix something else in the debian
installer.
> > Of course, for mini.iso it's a completely different story, enabling
> > unreleased could be useful there, help welcome on this.
> If you think the script's a good place for it,
Which script?
> I'll whip something up and send it to debian-boot@, with your
> blessing.
Getting non-release mini.iso use the unreleased series would probably be
useful not only for the hurd port, yes.
Samuel
Reply to: