init.d does some other stuff that basically sets method=efi on things that "look like" ESPs. That seems faintly risky to me but I guess it doesn't hurt us when we EFI boot so eh.
It would be simpler code to remove the creation and all checks for /var/lib/partman/efi but also eh.
On Tue, 26 May 2020, 22:46 Michael Hudson-Doyle, <email address hidden>
wrote:
> init.d does some other stuff that basically sets method=efi on things that
> "look like" ESPs. That seems faintly risky to me but I guess it doesn't
> hurt us when we EFI boot so eh.
>
> It would be simpler code to remove the creation and all checks for
> /var/lib/partman/efi but also eh.
>
New changelog entries:
* check.d/efi: Make sure we block on a missing EFI partition no matter what
architecture, not just for ia64. One could attempt to install on EFI x86
and will need an ESP to be able to install GRUB. (LP: #1803031)
* debian/partman-efi.templates: Make the no_efi template clearer, so that it
clearly explains why an EFI System partition is important.
New changelog entries:
* Do not ask the user about forcing UEFI installation if there are BIOS
compatibility mode installed systems as those will still be bootable from
grub if we proceed in UEFI mode. (LP: #1668148)
* debian/partman-efi.templates: remove the partman-efi/non_efi_system
template as we no longer need it.
New changelog entries:
* Resynchronise with Debian. Remaining changes:
- Remove efi-modules dependency; it seems to be built into Ubuntu
kernels now.
- Require partman-base >= 129 to support a null value for the name.
- Only force an EFI system partition on ia64.
- Allow x86/macs booted in legacy mode to use EFI System partitions.
I'd go one step further and drop the check in commit.d too.