I'm sorry to say it but it looks a lot like m68k@ is due for retirement in Gentoo. It seems like the last stage3 for m68k was.. 2013 [0]? There's currently a keywording bug of increasing importance (as we move to libxcrypt) which hasn't received any attention. I don't have access to any hardware but I think Chewi does. I know kent\n tried for a while to update from the last stage3 using qemu-user, but not sure how far he got in the end. While I'm not actually keen on losing support for something in Gentoo, I also like being up front about the state of affairs. Right now, I think it might be misleading to have it as a keyword at all. We certainly dropped sh@ over less, anyway. I propose, notwithstanding feedback from e.g. Chewi, that we make moves to drop the m68k keywords and profiles entirely. [0] https://gentoo.osuosl.org/experimental/m68k/
I'm not sure why exactly Kent was finding it so hard, except that he insisted on running all the tests. I don't think we need to set the bar that high? I do have real hardware (in the office that I haven't been for in 18 months) but it's far too slow to build stages on. I know that QEMU recently added a new much more powerful m68k machine type for situations like this one. I'll see what I can do by the end of the month.
IIRC we've dropped SH because nobody had any hardware left. If Chewi's going to work on it, I don't mind keeping it exp.
Despite I am having hardware and even put some considerable efforts in the past I am wondering if we have any benefits of having this arch even in exp. It has no practical use at all, no docs, no stages, no manpower to support it. I think we give our users false impression of 68k being supported (just because it is supported in kernel it does not mean it is supported in the distro).
(In reply to Michał Górny from comment #2) > IIRC we've dropped SH because nobody had any hardware left. If Chewi's > going to work on it, I don't mind keeping it exp. Yes, my DreamCast had gone. I later restored the unit but I am not enthusiastic of reviving the sh port.
(In reply to Mikle Kolyada from comment #3) > Despite I am having hardware and even put some considerable efforts in the > past I am wondering if we have any benefits of having this arch even in exp. > It has no practical use at all, no docs, no stages, no manpower to support > it. I think we give our users false impression of 68k being supported (just > because it is supported in kernel it does not mean it is supported in the > distro). Give me a chance, I'm working on this right now. I've got my existing m68k system running under the new QEMU machine with 3GB RAM already. Now I'm updating it. Even in the state that it's in, I'm aware of users in the forum trying this stuff with some success. I consider SuperH to be a rather different case. The hardware was much more obscure, with the Dreamcast being the only example I know of, off the top of my head. m68k was used in very many different machines that slightly are more practical for running Linux on. There are also significant ongoing efforts within the community, including getting LLVM and Rust working on it, as well as the QEMU work.
Status update. I have successfully built a stage3 and done a bunch of keywording. dilfridge is keen to add m68k to the set of stages that are automatically built via qemu-user, but his initial attempt failed because I accidentally missed zstd in my keywording. Unfortunately it blows up quite badly and it's a hard dependency of Portage, despite only being used for binpkgs. This certainly isn't the only package to fail tests but nothing else has failed this badly. I'll do what I can to figure it out.
The autobuilds are running fine now. Upload is only blocked by bug 810817 (i.e., infra, sigh).
m68k is back in business! We have autobuilt stages and I am doing lots of keywording. I'll try to keep up the momentum so that we can maybe promote the arch back to "dev" status.
