IRC channel logs
2025-01-13.log
back to list of logs
<mid-kid>Not as huge a rewrite as a last time, but redid a few things. <mid-kid>Now relying more on portage's dependency resolution to figure things out, slightly less hardcoding. <mid-kid>And the post-cross compilation steps are shorter too. <mid-kid>Only thing I didn't manage is upgrading to gcc 14. <stikonas>mid-kid: nice! much shorter than your 1st version <stikonas>mid-kid: by the way, why are we unpacking squashfs tree rather than tar.xz? <mid-kid>Thanks! I had already shortened it significantly in 2024.8, I just hope this is more future-proof. <mid-kid>I use the squashfs tree because that one's kept in the mirrors forever, whereas the tar.xz tree vanishes in a month. <rkeene>stikonas, All my changes should be merged in if you want to check it out <stikonas>sure, I'll update stage0-posix, though might not be today <aggi>continued rebasing against linux-2.4; need a filesystem with 64bit timestamps with a 32bit posix linux-2.x abi <aggi>won't be easy, since internal data structures changed significantly between linux-2.6 and linux-2.4, nilfs2 was introduced with 2.6 series which i can't support yet with tinycc <aggi>already pulled into some of the ancient grsecurity patches too, at least ASLR seems to work with x86_32 to shield against trouble <aggi>nilfs2 backporting is giving me a headache, needs a careful review, if and which filesystem could be prepared for 64bit timestamps <aggi>too tried a few of various recent linux-live.iso which ship with some efi-compat boot loader and a compliant MBR with active partition marked <aggi>interestingly, the ancient thinkpad t40 PC-BIOS doesn't appreciate such hybrid iso when those contain an efi-compat boot-loader <aggi>seems intel wrecked UEFI/CSM comaptibility from both sides, uefi firmware itself and however the legacy compat boot-loaders are implemented nowadays, doesn't work reliably with an IBM PC BIOS <aggi>with usb-sticks in particular <aggi>seems nilfs2 won't work with linux-2.4; giving up on this one radix-tree.h <aggi>with nilfs2 support it's two criteria, it's specific features themselves, and 64bit timestamps <aggi>because, if i hacked ext2 for 64bit time, then this probably causes breakage in userspace with 32bit posix, while nilfs would not <aggi>it's a little time remaining until hitting year2038 issues <aggi>and i have another suspicion, why intel seems in a hurry to phase out x86/32, including pc-bios boot <aggi>since whatever is hiding inside IntelME, if it's a 32bit posix i am not aware of any filesystem for year2038