IRC channel logs

<rickmasters>Googulator: I'm showing checksum issue starting with this commit: https://github.com/fosslinux/live-bootstrap/commit/a76fac4fa796fc17ab385595425dbbf1e4e1d7a0

<rickmasters>Googulator: some env variables are different after that so maybe that's the cause. I don't know.

<rickmasters>Googulator: what's strange is that several environmental variables on kernel bootstrap appear to match chroot better so not sure what's going on.

<rickmasters>fossy: somehow, the populate_device_nodes has fixed the checksum difference for --qemu with --kernel,

<rickmasters>fossy: manually doing a mkdir / mount resulted in a checksum difference for tar later on. I don't understand it but its working better.

<rickmasters>fossy: of course populate_device_nodes fixed kexec as well

<rickmasters>I'm praying it will fix the python checksum issue for kernel bootstrap but that might be asking for too much.

<Googulator>interesting, bisect here pointed to https://github.com/fosslinux/live-bootstrap/commit/8341c4e0894992eb03ded4a2b63f3db5baa02496

<rickmasters>Googulator: the farthest back I tested was three commits after that, d5008f9, which was successful.

<rickmasters>So with populate_device_nodes, qemu with --kernel linux fails on python 3.1.5 checksum just like kernel bootstrap.

<rickmasters>That's a big improvement though so I'll submit that PR as soon as I finish testing to make sure chroot and bwrap are unaffected.

<rickmasters>But it looks like populate_device_nodes hasn't fixed the python checksum issue so work remains on that.

<pder>Googulator: With latest master (754d967b2994b0a093c57) Fiwix still wont boot for me on this machine. I have tried booting another 32 bit linux and have gathered the e820 info from dmesg.

<pder>I downloaded a i686 version of debian live and ran "qemu-system-x86_64 -m 4096M -boot d -cdrom debian-live-11.8.0-i386-standard.iso -net nic -net user,hostfwd=tcp::2222-:22"

<pder>Here is the output from dmesg on the guest VM: https://paste.debian.net/1303862/

<pder>Please ignore that last link since I did not have kvm enabled in qemu: The following is the dmesg output from the kvm enabled 32 bit linux: https://paste.debian.net/1303863/

<GoogulatorMobile>Re-adding populate_device_nodes after finalize_fhs fixed the checksum issue for me.

<GoogulatorMobile>Filed a PR to fix the regression, although this isn't really my preferred fix, and we should probably investigate more.

<GoogulatorMobile>But regressions take priority.

<fossy>GoogulatorMobile: how very odd

<fossy>i would have said /proc, but /proc is mounted by finalize_fhs!

<fossy>is /dev/shm needed for semaphores, perhapse?

<fossy> https://man7.org/linux/man-pages/man7/sem_overview.7.html seems to allude so, maybe adding a /dev/shm mount to finalize_fhs would work too...

<Googulator76>fossy: testing that fix now

<Googulator76>also, autogen-bootstrap has had a new tag rolled out, so I updated the PR

<stikonas>just fixed a couple of bugs in posix-runner, now fork/execve seems to be working a bit better better

<stikonas>in fact on baremetal system mescc-tools-mini-kaem.kaem script from stage0-posix-AMD64 ran to completion

<stikonas>though I'm getting random hangs on qemu...

IRC channel logs

2024-01-12.log