IRC channel logs
2025-06-21.log
back to list of logs
<lanodan>Yeah, quite makes me hope it won't be needed for something like live-bootstrap <stikonas>well, it's not needed for bootstrapping gcc <stikonas>in principle it could be added as optional branch, but it's a lot of work to add it (need to buid a lot of different versions of clang...) <oriansj>honestly, every language feels like it is trying to make itself unbootstrappable. <agg1>noticed a peculiarity last week: bash.static compiled with tinycc segfaults and or mis-behaves <agg1>tested bash-3.x 4.x 5.x, either segfault or mis-behaviour when bash.static handles brace-expansion <agg1>i got a complete tinycc compiled system up and running, too salvaged a python.static with it, to retain gentoo/portage tooling <agg1>and then, got hit by surprise, it's not trivial to compile a bash.static without gcc or llvm/clang involved <agg1>fun one, bashism inside portage/eclasses prevents me from swapping bash with oksh <agg1>i've activated a tiny test-site documenting a few things of interest to bootstrappable.org, and you're cited rather often: <Mikaku>agg1: not sure if your segfaults are related to this but just in case don't forget to add '--without-bash-malloc' when you run ./configure during Bash build <stikonas>oriansj: indeed. I just had a thought. Given current zig "bootstapping" strategy where they commit web assembly blob into the repo, eventually they'll also have issues with git repo size... <ekaitz>stikonas: yup the repo is already huge <ekaitz>stikonas: I talked with andrew kelley about it not that long ago, and I started with an idea of his: writing a zig interpreter for the simplest case that could compile to that webassembly file <ekaitz>i even started writing it but that uncovered some issues in the guile standard library and fixed those instead <ekaitz>and then got overwhelmed by a lot of work and I couldn't continue <stikonas>well, yeah, there is always more work than any of us could do <ekaitz>andrew wanted to write that interpreter in Lua, and I started in Guile instead <ekaitz>from Zig they are not going to do it, but they support efforts to try make it happen <stikonas>well, probably still a lot of work, there is no way around having to write 2nd implementation of the language <stikonas>if you don't want very long bootstrap chain <ekaitz>yeah, but they had that in C++ before. I think one written in Guile/Lua/... would be way easier to make <ekaitz>specially if it's done for the simplest case <stikonas>yeah, even slow interpreter is probably faster than 50 or so steps (that are not too fast either)