IRC channel logs

2019-12-12.log

back to list of logs

<janneke>civodul: ooh, nice!
*janneke types --host=i686-guix-linux-gnu
<civodul>:-)
*civodul -> zZz
<civodul>night!
<civodul>42deeaea0ebbc2ccb54b8cf85a0660158835504490181e516623579a46b682c0 /gnu/store/niv24i583siranb6ydphx991dkvj8x8g-mes-boot-0.19
<civodul>from an initrd build \o/
<janneke>civodul: whoa!
<janneke>civodul: the 42*c0 is bin/mes, or what?
<civodul>janneke: the full store item, as in: guix hash -r -f hex /gnu/store/niv24i583siranb6ydphx991dkvj8x8g-mes-boot-0.19
<janneke>beautiful!
<civodul>the nice thing was to do it full-scheme
<civodul>with Guile 3, i wonder how feasible it would be to get rid of libgcrypt, for instance
<civodul>on Guile 2.0, the pure-scheme sha256 is noticeably slower than libgcrypt
<civodul>maybe 10x or so
<janneke>ah, yes; that's a great perspective
<janneke>things are starting to fall in place
<civodul>Scheme Machines are coming! :-)
*janneke just produced a 9*fb mes-mescc with tcc-built MES+MESCC and also with a GUILE+MESCC build
<janneke>i wonder what vagrant thinks of that kind of DDC
<civodul>yeah, that's very much DDC this time i think!
<civodul>at least for the TCC bit
<janneke>the fact that our C compiler is really a scheme program, i started to wonder how much the C compiler of MES or Guile matters...
<civodul>true
<civodul>hmm
<civodul>we'd need to reread Wheeler's thesis
<civodul>i guess it's really focusing on compilers written in the language they compile
<janneke>yes
<janneke>i think vagrant has been at least reading parts of it, lately
<janneke>*vagrantc
<janneke>civodul: did you build mes-0.19 on plain guix too; do they match yet?
<civodul>janneke: only mes-boot
<civodul>it takes ~5 GiB of RAM though :-)
<janneke>yeah, mes-boot-0.19 -- it matches! \o/ \o/
<janneke>hmm, 5 GiB seems is a bit much, not a problem for now -- this is amazing!
<janneke>hannes and vagrantc will appreciate this too
<civodul>janneke: i've pushed a section on this in guix-artwork.git, feedback welcome!
<civodul>i'd send the link to hannes but cgit is currently down on Savannah :-/
<janneke>civodul: extreme bootstrapping, beautiful!
<janneke>i'm planning to release wip-hurd with 0.22, fits well into that
<civodul>oh oh!
<civodul>neat
<janneke>i could not help notice that the `#bootstrapping' section is a bit empty still ;)
<janneke>so that runs mes, the scheme interpreter and mescc until it wants to fork/exec
<janneke>i was planning for 0.22 to be the grand `mes-m2 merge/build from 500bytes' release, but i found this new theme: build reproducible on debian/nixos/arch, port to hurd/freebsd theme that's pretty nice
<civodul>yup, both sound good
<civodul>one thing at a time, i guess
*janneke goes afk for a bit
<civodul>hannes: https://git.savannah.gnu.org/cgit/guix/guix-artwork.git/tree/website/drafts/reproducible-build-summit-2019.md
***ng0_ is now known as ng0
<gio>The "kernel extreme boostrapping" is basically what I am doing with asmc.
<gio>And the "initrd extreme boostrapping" is very similar to what I am doing with nbs, except nbs is not guix-specific.
<gio>Granted, both are still far from being complete.
<gio>(also, nbs is not Scheme-based, but C-and-shell-based)
<civodul>gio: oh, i wasn't aware of that
<civodul>do you have pointers to your work?
<civodul>we could mention it
<gio>civodul: Repositories are at https://gitlab.com/giomasce/{asmc,nbs}
<gio>I presented about them here: https://debconf19.debconf.org/talks/152-bootstrappable-debian-bof/
<civodul>neat, thanks gio!
<gio>(although that presentation is Debian-oriented, of course)
<gio>But so far nothing there is Debian- or distribution-specific.
<civodul>cool, looks like we'll have a lot to talk about and share :-)
<gio>I also have some preliminary work on how to compile Linux with tinycc and boot it from asmc, but it is not functional yet. I'd say the biggest part still to do is working around the fact that Linux requires a linker and a loder far more advanced than what tinycc offers.
<gio>I already managed to do a similar trick with iPXE, but Linux is still to do.
<janneke>gio: wow, i completely missed that you had already presented your work
<gio>Of course I am interested to know what you (plural "you", I mean, everyone in the chan) think about my presentation.
<civodul>gio: sure, i'll take a look
<gio>civodul: Thanks!
<janneke>"debian is not bootstrappable" ;-)
*janneke is enjoying the talk by gio
<janneke>gio: very nicely done, i especially liked your debian perspective and how you introduce other efforts
<janneke>gio: the talk helped me to understand what you're aiming at, i didn't really get that before
<janneke>good job!
<gio>janneke: Wow, that's great! I like when presentations enable others to understand what you are doing and why, since that's precisely why we have presentations! :-)
<fosslinux>janneke: how big is the seed for your initrd bootstrap?
<fosslinux>and how far are we from the 500 byte release; is that still waiting on oriansj's mes-m2?
<janneke>gio: yes, it's also nice to see your enthousiasm, working on similar things
<janneke>fosslinux: we haven't done / are not yet attempting an initrd bootstrap
<fosslinux>oo ok
<janneke>what civodul just did, was drafted a single guix package build in the initrd
<fosslinux>right
<janneke>he managed to build `mes-boot-0.19' in the initrd
<janneke>and compute and print the hash
<janneke>that means for guix that the build daemon has been removed from the trusted computing base, i think
<fosslinux>ohhh i get it
<janneke>we will probably try to do this for interesting nodes in the graph
<xentrac>nice!
<civodul>gio: i enjoyed https://debconf19.debconf.org/talks/152-bootstrappable-debian-bof/
<civodul>particularly the 4 life stages metaphor!
<civodul>i think it works well, and it helps clarify what the various approaches are about
<janneke>yes, i agree
<civodul>i realize i had pretty much been disregarding stage #1 until now
<civodul>well, until last week ;-)
<janneke>yes!
<civodul>anyway, great talk, and it's good to see increasing mindshare in this space
<xentrac>I don't suppose there's a transcript available yet, is there?
<janneke>it always felt like in "bootstrapping" we are solving some related but different problems, and this gives a nice angle on that, a way to categorise and think about it
<xentrac>yeah, there are various different reasons bootstrapping is interesting, which sometimes leads us to talk at cross-purposes :)
<civodul>janneke: right, i think i used to view stage #1 as something different from what you were looking at
<janneke>civodul: yes, i'm very happy with our progress
<oriansj>civodul: actually we have a plan to remove Linux from the TCB and infact remove the concept of bios, firmware and microcode from the bootstrap entirely.