IRC channel logs

<ekaitz> https://notgull.net/announcing-dozer/ amazing

<ekaitz>rust -> ocaml -> guile -> C hehe

<fossy>lol, i was quite impressed by that

<ludocode>There's a discussion on Hacker News about it: https://news.ycombinator.com/item?id=41351446

<ludocode>Another Rust compiler (especially a boostrappable one) is probably a good thing, but I'm skeptical about whether it will be successful, given the state of other Rust compilers...

<ludocode>mrustc has been stuck on 1.54 for years. gccrs has been under development on and off for like 8 years and still can't compile Hello World.

<pabs3>from #lobsters: Bootstrap a Linux system from a 512 byte binary seed with 170 steps [linux programming compilers] (andyc) https://lobste.rs/s/obokni

<pabs3>another Dozer discussion: https://lobste.rs/s/fekehz

<pabs3>TIL a Lisp interpreter written in Awk https://github.com/darius/awklisp

<nikolar>That's cool

<fossy>hn also: https://news.ycombinator.com/item?id=41352055

<fossy>largest increase in stargazers for live-bootstrap in a day from these two, i believe

<Mikaku>fossy: Fiwix web site also experienced a notable increase in stargazers with: https://lwn.net/Articles/983340/ probably live-bootstrap as well

<aggi>a little more patching to tcc, and it can be utilized for bootstrapping in a cross-compilation setup

<aggi>besides, i noticed a slight iregularity when re-confirming gcc-bootstrap with tcc (using a patched gentoo ebuild/eclass and crossdev)

<aggi>that is, the gcc makefile.am do not support it when CBUILD!=CHOST (it is overriden CC|LD|... which isn't trivial to disect)

<aggi>otherwise, i had seen gcc bootstrap succeed when CBUILD=CHOST=CTARGET, then CC=tcc is corretly applied by the gnu buildsystem

<aggi>anyway, i'll test a i386-tcc cross-toolchain setup on an aarch32 buildhost today

<aggi>the cross-compilation setup revealed a few iregularities for bootstrapping already

<aggi>although i am not doing it with bootstrappable scripting, but a patched gentoo overlay and a patched tcc-toolchain

<aggi>a major concern remaining: gnu autotools and gentoo/portage bdeps are not self-hosting with tcc-toolchain

<aggi>i had wasted yet another night for example, digging into perl buildsystem, to support static-linking

<aggi>i merely couldn't decide yet, if it's less efforts to fully re-integrate all makefile.am and the entire packaging since when i lost the required tooling

<aggi>or if it was easier to repair python and perl for tcc-toolchain support and static linking

<aggi>in any case, i could confirm a complete set of ~500 ebuilds already, a fully capable userspace (including ffmpeg,mplayer,mutt mailer etc. etc), that doesn't need any binutils nor gcc at all

<aggi>as far as kernel is concerned, a production-ready kernel image for this is necessary, that can't be emitted with tcc-toolchain yet

<aggi>but i got a few ideas and options, and i may succeed emitting a devdrop for i486-tcc-linux-musl.iso this year

IRC channel logs

2024-08-26.log