IRC channel logs

<stikonas>fossy: hmm, given how long CI is taking on github actions, maybe we should not run everything?

<stikonas>splitting into more and more parts looks a bit fragile

<stikonas>fossy: also I think qemu mode (with kernel) is broken

<stikonas>at the beginning of sysc after installing packages I got mv: cannot overwrite directory `/distfiles'

<pabs3>LWN reposted the Guix news: https://lwn.net/Articles/930650/

<fossy>stikonas[m]: i was thinking the same thing, however that needs a bit more intelligent logic

<fossy>i just want to get this out as a stopgap solution then i'll work on something proper

<emilytrau[m]>ACTION uploaded an image: (4007KiB) < https://libera.ems.host/_matrix/media/v3/download/matrix.org/IfHJNtLeLkxhnTUOhpSdEDCz/nixpkgs-minimal-bootstrap-graph.png >

<emilytrau[m]>The dependency graph for nixpkgs stage0-posix -> tcc+mes 😍

<emilytrau[m]>ACTION uploaded an image: (132KiB) < https://libera.ems.host/_matrix/media/v3/download/matrix.org/NiqJaMvupTmaLedqXnWdDEUY/image.png >

<emilytrau[m]>such a beautiful mess when you put it next to guix's lmao

<civodul>emilytrau[m]: there's quite a few more nodes here :-)

<civodul>it's a derivation graph though; the Guix one above is a package graph, so it's more concise

<civodul>anyway, thumbs up!!

<janneke>emilytrau[m]: very nice, when you have tcc the hardest part is done!

<stikonas[m]>emilytrau: we have a bit outdated graph for live-bootstrap too

<stikonas[m]> https://stikonas.eu/files/bootstrap/live-bootstrap.svg

<stikonas[m]>This is handwritten, so a bit harder to keep up to date and complete...

<janneke>ACTION cleans up wip-riscv and adds a (broken) kaem -f kaem.riscv64: https://lists.gnu.org/archive/html/bug-mes/2023-05/msg00000.html

<stikonas[m]>janneke: yes, I don't expect to be big issues in building mes-m2 for riscv. I have done it once when i worked on stage0-posix-riscv

<stikonas[m]>Though that was probably still with bootstrap mode...

<stikonas[m]>But I couldn't run it at all in qemu...

<stikonas[m]>Anyway it looks like we first need new release of M2-Planet...

<janneke>stikonas[m]: yes, i thought it would be goot for Mes to support an M2-Planet build ootb, even if it crashes for now

<stikonas[m]>janneke: by the way your riscv build of mes is expected to segfault on qemu binfmt

<stikonas[m]>qemu only allocated 16 MiB of brk space

<stikonas[m]>They expect most people to use mmap

<stikonas[m]>Or maybe 32 MiB now but still not enough for mes (https://lists.gnu.org/archive/html/qemu-devel/2022-01/msg02958.html)

<janneke>stikonas[m]: thanks!

<janneke>ACTION is making a little progres, had some typos in instruction set -- do'h

<stikonas[m]>I was initially hitting this limit in hex2 or M0 on riscv...

<stikonas[m]>So I had to optimise string storage a bit

<janneke>*woot* mes-m2 runs on riscv64

<janneke>some cleanup todo

<h01ger>wheehooo

<stikonas[m]>janneke: why don't you use stage0 style riscv M1 macros?

<janneke>stikonas[m]: no idea, i just used wat laanwj used

<stikonas[m]>Current ones are way more complicated

<stikonas[m]> https://github.com/oriansj/M2libc/blob/main/riscv64/riscv64_defs.M1 can be used to encode everything from rv64m

<janneke>stikonas[m]: ah right, instructions can be composed now

<janneke>mes needs to be cleaned-up in that respect, i guess

<stikonas[m]>Only riscv instructions can be composed

<janneke>ah, ok

<janneke>ACTION is afk for a bit

<stikonas>janneke: is something like that fine https://paste.debian.net/1279218/ ?

<stikonas>macros are slightly more obscure but they can be constructed from individual bits

<stikonas>or do you prefer mes original macros?

<stikonas>(the main advantage of stage0-posix riscv64 macros is that anything in RV32I, RV64I, RV32M and RV64M sets can be encoded

<stikonas>(without any additions to defs file)

IRC channel logs

2023-05-03.log