“Also if you want to configure PAM to do something special such as send a slack message to some channel whenever someone logs into a machine or runs a sudo command and that option is not already in the pam options in nixpkgs, you’re basically doomed because pam is not very configurable on NixOS.“
@xerz Someone reported that badwolf works on x86_32 (as should be) and webkit-gtk is a rather common dependency, no idea how good it is though, JS engine could be slow.
At this point, while I love the workflow so much and it’s gotten me somewhere already, it’s starting to look like NixOS is constraining me a bit more than I need.
ftr, I just checked aaaaaaaaaaaaand yeah, the OS compiled indeed, it’s all 32 bit binaries, and if I try to run e.g. coreutils or Netsurf on my Arch, they run just fine
so I definitely borked the kernel with the custom flags, I just wonder if I can make Linux itself an exception
OK IT’S DONE AND IT WORKS, I CAN NOW USE PROPER 32 BITS AND CFLAGS EVERYWHERE :blobcatrainbow: :blobcatrainbow: :blobcatrainbow: :blobcatrainbow: :blobcatrainbow: :blobcatrainbow: :blobcatrainbow: :blobcatrainbow: :blobcatrainbow: :blobcatrainbow: :blobcatrainbow:
basically had to find out how to set nixpkgs with config.replaceStdenv ‒ apparently overlays don’t apply to every package and so I was getting ABI inconsistencies, so a lot of time was spent in a dead end (thx @K900 !)
Not sure if I should further adopt Clang and/or some form of linking optimization (LTO, gold, whatevers)… this is already taking an eternity to build for what should be very trivial comparisons :blobcatsweats:
Also, -Oz could be fun to try regardless of compiler, but I’m betting the most likely thing is that it will make my binaries way slower on old computers, so… probably not worth the hours upon hours of compile time?
OK waiting for failed optimizations is taking away the fun out of making an OS, I’ll leave that for last and maybe wait to save up money for some hours on a build farm or something
OK so I'm trying to boot on a weird i5-480M+Radeon 5400… and when it reaches X11 it just stops working. Unless I go nomodeset, then X11 can't find a display.
gosh do I love playing around with Quecto, I can do a query as large as grep /root $(find / -type f -name '*.mount') within mere milliseconds :blobcatmelt: 💗
like, it’s making my life terrible because I cannot configure a default path for nix --store, but I can also do things like systemd-analyze plot | nix run nixpkgs#feh -- - and it looks supercool :blobcatfluffowo:
ok so now I have a brand new issue: I cannot boot from ext4 drives somehow??????????????
like, it’s perfectly fine as soon as I add fsType = ext4, I made a diff of the derivations’ JSON to see if there was anything really new but uh… yeah it only tells me the hashes of the packages it has generated, and I don’t know how to look further
It should be the other way around, somehow btrfs is autodetected but not ext4?????????????
“ You can usually omit the file system type (fsType), since mount can usually detect the type and load the necessary kernel module automatically. However, if the file system is needed at early boot (in the initial ramdisk) and is not ext2, ext3 or ext4, then it’s best to specify fsType to ensure that the kernel module is available. “
furthermore, adding or removing ext2/ext3/ext4 from boot.initrd.supportedFilesystems doesn’t matter at all, since it all evaluates to the exact same derivations…
like, all I can guess from here is that it can't see the label without figuring out the filesystem first, and I guess it's trying to find the partition that matches without trying to figure out the filesystem first... which somehow works with Btrfs but not with ext4?????
so in essence and presuming this is the kernel's fault, at this point I should choose between either LABEL (deprecating ext4) or GUID/PARTLABEL (deprecating MBR)
ultimately, both tradeoffs suck and feel completely artificial, but one of them only affects existing hard disk drives with MS-DOS and alike, while the other affects the most common use cases for formatting...
so I ran this wonderful tool called nix-tree, very straightforward to use
so uh, apparently rav1e is required for libheif which is required for feh which for some goddamn reason is required by xsession???????????????????????????????? :blobcatgooglyheadache:
Add comment