Re: NetBSD-10_BETA gpt sysinst experience

[David Brownlee <abs%absd.org@localhost>]

On Thu, 12 Jan 2023 at 11:58, Martin Husemann <martin%duskware.de@localhost> wrote:
>
> On Wed, Jan 11, 2023 at 11:16:13PM +0000, David Brownlee wrote:
>
> > ** The EFI should probably not be "a"
> > ** Likewise having "/" as "b"
>
> As Robert explained, that letters are menu items, not partition letters.

Aaah! should have worked that one out. Too many years working with
disklabels. Might be worth a note in the surrounding text, but this
one is essentially a user problem my end

> > ** "c" and "d" should probably be reserved and not showing up as <swap> and /tmp
> > ** Filesystem details column is truncated with around 20 spare chars
> > on right (minor)
>
> Need more details for that, like your $TERM and the window (or framebuffer)
> size and the full content you expected (and maybe a picture of the screen).

Image at https://imgur.com/a/wjzvhgK

> > 2) Selecting install > wd0 > Preconfigured wedges
> >
> > Gives a fine display of the existing wedges (no way to proceed)
>
> This menu item is meant for semi-manual setup. Currently sysinst does not
> really support multiboot EFI setups (and this is non-trivial).
>
> So the idea is: you setup whatever you need for boot selection (copy the
> netbsd bootloader to the ESP and add proper variables/configuration
> in the EFI shell or whatever) and also (manually) create the NetBSD / ffs
> partition and (optionally) the NetBSD swap partition (both with proper
> partition types).
>
> Then you can use sysinst to install into the pre-created dkN for / (and it
> will magically find the swap partition and offer that).
>
> This is also the way to go if you have a more complex cgd/raid/lvm setup
> (e.g. / on cgd). You prepare everything manually and the use this menu
> item to populate / and do the non-partitioning part of the setup.

That seems very reasonable.

> > 2) Selecting install > wd0 > Extended
>
> This is a still quite shaky area and not fully tested. I spend some time
> fixing more ways through it every now and then, but it is not high priority
> (and there are a million ways through it).

Thanks for that - as I mentioned in the original email this looks like
a very interesting feature to have.

Are there any changes in -current sysinst that would make it more
useful for me to test against that?

Should I send-pr for the below?

- The second and subsequent time I enter the extended partition
manager the display overwrites with kernel messages as it recreates
wedges - is this triggered by sysinst, and if so might it be possible
to add a short delay before showing the page, to allow the messages to
(usually) be finished before the page is shown?

- When using the extended partition manager to add a partition to an
existing gpt setup, the partition appears to be created, but the
newfs/mount options appear to be ignored, leaving sysinst installing
into the install image and aborting with disk full

- When using the extended partition manager to add a partition to an
existing gpt setup, the partition appears to be created with an all
zero uuid

Thanks