git bisect [1] should be the perfect way to find the culprit here. Are you up for it?

[1] https://wiki.archlinux.org/title/Bisecting_bugs_with_Git

https://gitlab.archlinux.org/archlinux/packaging/packages/grub/-/commit/beee9df4ae2e3b2150d4e54b3825064e77661cb4

stops working

Sorry if I was unclear. I meant git bisect the *upstream* repo to find the offending upstream commit.

if [ "x${GRUB_DISABLE_UUID}" = "xtrue" ]; then

if [ "${GRUB_DISABLE_UUID}" = "true" ]; then

schould it be "true" insted of "xtrue"?? removed 2 x, ArchLinux commit in PKG.

Just a short side note... The package with version `2:2.06.r557.gac560966d-1` failed to boot on *ALL* EFI systems. So let's ignore that.

But to get it right... The system fails to boot with `2.06.r499.ge67a551a4-2`, but `2-2.06.r499.ge67a551a4-1` was ok? That's kind of crazy, as both are built from same source (with updated unifont and configuration dropings on top).

Did you try recent packages? Currently we have `2:2.12rc1-1` in repositories.

i tried every version, need to chroot everytime to downgrade

greater 2.06.r499.ge67a551a4-2, it works > not >=

tested a few things, its a upstream bug.

The "x" does not change anything if you add or remove it on both sides. 😉
Some projects add it to make sure no empty string is used in the comparison.

can be closed and marked as upstream

I just hit this bug trying to reinstall a Arch Linux system on a HP Pro Tablet 408 G1. Are we saying this is a upstream issue as in a bug with the GRUB code?

Perhaps this can help.

On the install pendrive I replaced the file EFI/BOOT/bootia32.efi with that found at

https://github.com/jfwells/linux-asus-t100ta/blob/master/boot

And with that I was able to boot correctly.

Best regards.

Maurizio.

With the downloaded module however I had a little problem, since the loader was not recognizing the '[' command present in some script.
So I've taken the ORIGINAL Grub sources, compiled them, created the efi module, copied it directly in my boot location and then booted. It works.

I was inspired by the mentioned site:

https://github.com/jfwells/linux-asus-t100ta/blob/master/boot/readme.md

I tried to download its sources and follow the instructions but I got several errors.

So I followed the instructions using instead the original sources of grub 2-06:

Downloaded the package from https://ftp.gnu.org/gnu/grub/

tar -xvf grub-2.06.tar.gz

cd grub-2.06

(the autogen command is not necessary)

export EFI_ARCH=i386

./configure --with-platform=efi --target=${EFI_ARCH} --program-prefix=""

make


The compilation went perfectly without even the slightest warning. Then :

cd grub-core

Then I used a slightly modified command to remove some module, add another (test) and specify a different prefix directory /grub instead of /boot/grub
(probably i could remove even more modules ...)
I neeeded to add the test module otherwise the '[' command was not recognized in the scripts.

Original command :

../grub-mkimage -d . -o bootia32.efi -O i386-efi -p /boot/grub ntfs hfs appleldr boot cat efi_gop efi_uga elf fat hfsplus iso9660 linux keylayouts memdisk minicmd part_apple ext2 extcmd xfs xnu part_bsd part_gpt search search_fs_file chain btrfs loadbios loadenv lvm minix minix2 reiserfs memrw mmap msdospart scsi loopback normal configfile gzio all_video efi_gop efi_uga gfxterm gettext echo boot chain eval

My command :

../grub-mkimage -d . -o bootia32.efi -O i386-efi -p /grub boot cat efi_gop efi_uga elf fat linux keylayouts memdisk minicmd part_apple ext2 extcmd xfs xnu part_bsd part_gpt search search_fs_file chain btrfs loadbios loadenv lvm minix minix2 reiserfs memrw mmap msdospart scsi loopback normal configfile gzio all_video efi_gop efi_uga gfxterm gettext echo boot chain eval test

Then I took the bootia32.efi module, copied it on the target machine in the EFI/boot/arch directory, and booted successfully.

Best regards.
Maurizio

I have since moved on to using systemd-boot after this bug was fixed. https://bugs.archlinux.org/task/79531