I'd like to bump quantal from grub2 1.99 to 2.00.
Firstly, I apologise for this being relatively late. For one reason or another I was a bit short on motivation to get it sorted out earlier, not to mention short on time; but neither of those are a problem any more, this is now a very high priority for me, and I can devote considerable attention to sorting out any regressions.
GRUB 2.00 is a significant release, and contains quite a number of feature changes. It is worth pointing out that we have backported many of these to 1.99 or originated some of the patches (particularly various filesystems, submenus, EDID video mode selection, and grub-mount), and by now our 1.99 packaging is a Frankenstein collection of backports. Between that and being a year behind, upstream is pretty much uninterested in going anywhere near it by now, and my mailbox is full of bugs where people report that 2.00 fixed things for them. I'm keen to bring us up to something more current and less extensively patched. Furthermore, there are several features that are important and impractical to backport. These include:
* USB EHCI (some devices aren't accessible by 1.99 due to the lack of this)
* EFI serial
* Network stack for (among other platforms) EFI, which should permit EFI netbooting (an important feature requested by some partners for 12.10)
* LUKS support
* A gfxmenu theme, which should help to eliminate the reasons people feel they need to use BURG (I don't know if I'll enable this by default in quantal, but it will be available)
* GDB stub and support script, including backtrace on crash on i386
* Faster device scanning
Here's the debian/changelog fragment describing the new upstream release:
* New upstream release.
- Add LUKS and GELI encrypted disk support (closes: #463107).
- Lazy scanning to avoid accessing devices which aren't really used.
This avoids boot delay due to slow device scanning (closes: #549905,
#550015, #550083, #564252, #595059, #632408).
- Don't override more informative errors when loading kernel/initrd
(closes: #551630).
- Support 4K-sector NTFS (closes: #567728).
- Unify grub-mkrescue interface on powerpc with that on other
architectures (closes: #570119).
- Fix infinite recursion in gettext when translation fails (closes:
#611537, #612454, #616487, #619618, #626853, #643608).
- Add more missing quotes to grub-mkconfig (closes: #612417).
- Import gnulib change to fix argp_help segfault with help filter
(closes: #612692).
- Support %1$d syntax in grub_printf (closes: #630647).
- Use write-combining MTRR to speed up video with buggy BIOSes (closes:
#630926).
- Remove multiboot header from PXE images to avoid confusing ipxe
(closes: #635877).
- Fix crash when attempting to install to a non-BIOS disk (closes:
#637208).
- Fix handling of grub-mkrescue --xorriso= option (closes: #646788).
- Use umask rather than chmod to create grub.cfg.new to avoid insecure
grub.cfg (closes: #654599).
- Improve font installation logic (closes: #654645).
- Add grub-probe info documentation (closes: #666031).
- Don't crash on canonicalize_file_name failure in grub-probe (closes:
#677211).
I'd like to call out the lazy scanning and gettext infinite recursion items in particular. While I've had enough on my plate with going through the bugs in Debian to go through the (on average rather less helpful and not curated by upstream) bugs in Ubuntu, these fix a large number of delays and hard-to-diagnose crashes, in the latter case especially on complex RAID/LVM/etc. setups. I would be lying if I said I didn't expect there to be some regressions, but I think we can cover the worst of that with advance testing in Debian and in a PPA, and I should certainly be able to deal with anything that's serious and widespread.
This also fixes bug 1046429, which is important for UEFI clients.
Please see https://launchpad.net/~cjwatson/+archive/grub for build logs. My testing in Debian has been encouraging so far, only needing to fix a few comparatively trivial problems encountered in my plain, LVM, LVM+crypto, and RAID-1 test VMs. I've encountered one DM-RAID-based problem which still needs to be fixed.
I'll attach ChangeLog (warning, large!) and NEWS diffs.
Status changed to 'Confirmed' because the bug affects multiple users.