On 1/28/19 4:42 PM, Frank Scheiner wrote: >>> I believe the current d-i/yaboot-installer expects to install on a HFS >>> bootstrap (though I assume yaboot can also boot from FAT). If the >>> default bootstrap method is changed this would require changes to keep >>> it working. >> >> Yes, but GRUB works with HFS, doesn't it? > > Sure, did anybody question that?
Well, your arguments seemed to imply that to me, yes. >> Because there are newer upstream versions available for hfsprogs as the >> code is still maintained by Apple: >> >>> https://opensource.apple.com/source/hfs/ > > In an earlier message I linked a discussion on Archlinux about using > newer upstream versions (see bottom of [1] for details, but don't forget > that this is for x86_64 not for powerpc/ppc64, as is the code from Apple). > > [1]: https://aur.archlinux.org/packages/hfsprogs/ I don't think the code is in any way specific to x86_64, is it? Apple is not know for writing code that is not portable, in my experience. They are certainly keeping it portable enough so that it will work on their ARM platforms. We should definitely try the updated package from Arch and see if it works on PowerPC as well. If it does, we should update the Debian package with the changes from Arch. >> Furthermore, if I remember correctly, we need to stay with an HFS boot >> partition because apparently using FAT wouldn't work with QEMU, wouldn't >> it? If I remember correctly, your suggestion to use FAT for the boot >> partition was rejected in a recent discussion for this very reason. > > Oh, it's new to me that it was already rejected. That is an interesting > way how decisions are made here, to say the least. You are misreading me here. I am not saying it was rejected at all, I am merely saying that the idea was rejected by at least one party. No decision has been made whatsoever yet. If we can agree on using FAT as the boot partition instead of HFS and we find a way to get it working in QEMU as well, I don't see any problem with adopting your suggestion. > Up until now I thought there was a discussion mainly between two persons > (me and Mark) with some degree of misunderstanding on one side about > what effect a possible change will really have. Yes. And I have not suggested anything else. You know that I am happy to merge any patches if they have been tested to work and agreed on. I don't have any strong opinion on how things are to be implemented other than that I am reluctant to support old and unmaintained code solely for personal preferences. All I want is that we are using GRUB as the default bootloader on as many platforms as possible, simply because GRUB is supported and maintained upstream while both SILO and Yaboot are not. It's simply a matter of making sure we have something that works and is supported. > And I thought I've made it clear multiple times now that my changes > won't make it impossible to use Debian on emulated Power Macs in QEMU > (for the sake of the millions out there... ;-)). Pardon me, but I'm not sure what the current status with QEMU now is. Does it work with FAT boot partitions? If FAT boot partitions work with emulated as well as non-emulated hardware without any issues, I have no problems adopting them. I just want to avoid extra code paths and switches as that will make our code more complex and prone to breakage. > Partitioning and bootloader installation to enable the HFS bootstrap > method just have to be done manually after the changes are applied. I would like to have GRUB working out-of-the-box on any of the platforms where GRUB is actually supported by upstream. > In addition, with the changes applied, we would really only rely on currently > maintained > software for installation and could still keep and even fix the other stuff. What other stuff are you referring to? To summarize my points: - I do not want to work on Yaboot because it is old and unmaintained and has multiple issues that are not addressed by upstream and is also compatible with powerpc and ppc64 only meaning it cannot live in unstable but unreleased only meaning it requires 100% manual maintenance which is something I would like to avoid - I want to use GRUB on as many platforms as possible with as little as possible kludges and workarounds in the code as possible since the code is supposed to go into debian-installer and components and the debian-installer team will not be happy if we include workaronds and hacks for older, unsupported software unless it is absolutely necessary - I would like to update hfsprogs to the latest version possible and I would like to ask anyone with access to a Debian powerpc and ppc64 test installation to test the updated package from ArchLinux (v540); if that version works, we should use it to update the Debian package - I do not keep anyone from using the bootloader of their choice; please use whatever you want; but please understand that I personally do not want to invest any work and effort into these old bootloaders since my time and resources are limited and I rather want to work on stuff that is supported and maintained upstream Adrian -- .''`. John Paul Adrian Glaubitz : :' : Debian Developer - glaub...@debian.org `. `' Freie Universitaet Berlin - glaub...@physik.fu-berlin.de `- GPG: 62FF 8A75 84E0 2956 9546 0006 7426 3B37 F5B5 F913
