IRC channel logs

<Googulator>andydude: let me suggest the name "m3" for the C autoconf replacement :)

<Googulator>we already have m0, m1 & m2 early in the bootstrap chain

<Googulator>& then m4 is used by autoconf for macros; IIRC you plan to keep using m4 for the C replacement too

<Googulator>hence the C autoconf replacement kind of fits in the chain between m2 (& many more components), and m4

<stikonas>oriansj already tried to use m3 for something, though I don't think there was any code written

<stikonas>or maybe a bit but nothing used in production https://github.com/oriansj/M3-Meteoroid

<stikonas>I think the intention was to have binutils compatible tools

<stikonas>rather than what mescc-tools use

<andydude>Googulator: I was actually considering calling the whole thing "atbox" and the autoconf library "libatm4te" the automake library "libatmake"

<andydude>I don't want to replace m4, I like m4

<andydude>I also like the idea of choosing names from some similarity pool, like cat names, mountain names, etc.

<andydude>Like maybe "autocheese" and "automarinate"

<andydude>or "libnoperlconf", "libnoperlmake"

<fossy>hey stikonas, i think your forgejo has some problems, just thought i should let you know :) (500 on trying to access gentoo-bootstrap)

<fossy>thanks mid-kid for _ctypes + dynlinks for python

<stikonas>fossy: oops, let me take a look

<stikonas>fossy: should be fixed now

<matrix_bridge><cosinusoidally> I've been experimenting with running stage0-posix on very old Linux distros. Turns out there is a bug in M0_x86.hex2 that causes it to crash on 2.0 kernels and earlier. M0 expects ebp to be initialised to 0 but Linux 2.0 and earlier don't do that https://asm.sourceforge.net/articles/startup.html . I fixed that bug that then allowed stage0-posix to run to completion on a 2.0.37 kernel (under slackware...

<matrix_bridge>... 3.9 from 1999). stage0 also nearly runs to completion under earlier kernels, but I think it runs out of memory as kernels before 2.0.36 do not detect more that 64MB of RAM in qemu.

<matrix_bridge><cosinusoidally> Under slackware 2.0.0 (with an 1.0.9 kernel from 1994). stage0-posix fails on "# Phase-15 Build M2-Planet from M2-Planet #"

<matrix_bridge><cosinusoidally> I think it's probably running out of memory. I managed to get my tcc_bootstrap_alt to run to completion on 1.0.9 but I had to reduce the memory usage of this step https://github.com/cosinusoidally/tcc_bootstrap_alt/blob/master/tcc_10/build.kaem by reducing the number of times I rebuilt tcc (fix is not yet committed to the repo).

<matrix_bridge><cosinusoidally> getting old distros working can also be a bit fiddly. eg Linux 1.0.9 doesn't seem to be able to detect disks larger that 512MB (or somewhere between there and 1GB). So I ended up scratching my head for ages figuring out why Linux couldn't find my disk image.

<matrix_bridge><cosinusoidally> For slackware 3.4 to 3.9 I found zipslack to be handy, since no matter what I did I couldn't seem to get the boot floppies to work on any 3.x version of slackware. zipslack allows you to install slackware on a dos partition. I first installed freedos 1.3 unzipped zipslack and then used loadlin to boot into slackware.

<matrix_bridge><cosinusoidally> (all under qemu)

<matrix_bridge><Andrius Štikonas> cosinusoidally: nice catch. Hopefully PR is incoming

<matrix_bridge><cosinusoidally> yep I'll raise one some point

IRC channel logs

2024-08-04.log