New Grub version

[robert@…]

grub-1.95 was releases last October. It depends on the LZO compression library, and Bison, and there's no test suite. The website says this is the legacy branch. It's really too bad it doesn't use libbz2, or even libz.

[alexander@…]

Thanks for testing, but now I have no stable opinion on the boot loader selection for LFS (except for the fact that I will use lilo, no matter what is in the book). grub-1.95 is as different from old grub as lilo is, while lilo has a much more stable code base.

Maybe it is a good idea to move grub out of Chapter 6, and document all known alternatives (lilo, grub, grub2, gujin, and maybe even etherboot) in the separate chapter. If this is the way we choose, I take lilo and etherboot.

[bdubbs@…]

Providing alternatives to boot loaders would break with the current single thread of building a system that is in LFS now.

I would prefer to see a hint or at least a text file describing the different alternatives with advantages and disadvantages of each. This should include build dependencies (e.g. lilo requires nasm), commands, and install procedures.

With such data, we could then add the alternatives to BLFS and point users there for details.

[Thomas]

Replying to alexander@linuxfromscratch.org:

Maybe it is a good idea to move grub out of Chapter 6, and document all known alternatives (lilo, grub, grub2, gujin, and maybe even etherboot) in the separate chapter.

++

This is required when leaving the x86 terrain anyhow. On PPC, the bootloader cannot be GRUB but must be yaboot.

[Jeremy Huntwork]

Replying to Moody:

Replying to alexander@linuxfromscratch.org:

Maybe it is a good idea to move grub out of Chapter 6, and document all known alternatives (lilo, grub, grub2, gujin, and maybe even etherboot) in the separate chapter.

++

This is required when leaving the x86 terrain anyhow. On PPC, the bootloader cannot be GRUB but must be yaboot.

I'm intending to do this very soon with the jh branch, mostly as a way for people to 'taste' the idea before it's seriously considered for trunk. I'll be dropping in a framework for it, and then hopefully I'll get some help filling in the contents. :)

[alexander@…]

I have tested Gujin-2.2 and must say that it is unsuitable for the book in its current form.

• It crashed KVM
• It reports "unknown key pressed" when running in QEMU (but still boots the first kernel and its initramfs when I press F1)
• It is impossible to install Gujin into MBR without losing all data on the hard disk, because it explicitly asks to zero out the first 63 sectors of the hard drive, thus wiping out the partition table
• Upgrades are problematic, because the boot sequence is protected by checksums - and if you don't upgrade the MBR contents (thus losing all data), old MBR wil refuse booting new Gujin.
• By default, it adds LANG=en to the kernel command line (bug reported to authors). Kernel transforms this into the environment variable, but "en" is not a valid value for LANG.

GRUB-1.95 (unlike the current version) is completely unusable (read: refuses to install) with root on LVM and /boot on a small partition, but, according to the Changelog, this has been fixed in the CVS version. So, I would like to postpone the update of GRUB.

[bdubbs@…]

I suggest we close this ticket for now. This ticket is over two years old. There have been no commits to grub 2 in over a year. When we get to 64-bit issues, then another ticket to address non-grub alternatives should be opened.

[bdubbs@…]

Marking as duplicate to #2093.