IRC channel logs
2024-04-17.log
back to list of logs
<fossy>stikonas: i think theres a fair chance that some of the patches dropped when Googulator updated to 4.14 means that 64->32 is now possible <fossy>but it 100% used to be broken <stikonas>well, the main motivation for this is running live-bootstrap on UEFI systems before proper UEFI kernel bootstrap is ready <stikonas>or running live-bootstrap on non x86/x86_64 systems <fossy>in what sense do you mean "UEFI kernel bootstrap"? <stikonas>where presumably userspace bootstrap will be done well before kernel bootstrpa <stikonas>well, start with arbitrary x86_64 linux.efi binary <fossy>or at least, any other architectures are no-support <stikonas>and proceed with mostly userspace bootstrap <fossy>Googulator: is there actually a PR for nyacc bootstrap or just a commit? <stikonas>(possibly jumping into fiwix or bootstrapped linux) <Googulator>I rolled nyacc bootstrap into it since mes 0.26 was a prerequisite for it anyway <Googulator>indeed, the main reason for going 0.26 was nyacc bootstrap support <fossy>i'll spend a bit of time checking that this actually does 100% fix the nyacc problem, but at minimum this is a big step forward, thank you ^-^ <Googulator>Unless there are more pregens in nyacc, this should be a fix <fossy>i'm surprised that it was actually fairly straightforward, i had in my mind that there was a fair few more blockers to it but maybe not at all! <Googulator>When doing bwrap bootstrap, psyntax.pp and all known nyacc pregens are deleted before we attempt regeneration <stikonas>fossy: well, there was some work done by other people <Googulator>After upgrading to 0.26 and adding samplet's code, there was really just one issue <fossy>yeah, the patch by samplet made a big difference <Googulator>so I just adjusted samplet's code to regenerate in place, rather than with .new suffixed filenames <Googulator>and luckily it does work, overwriting the pregens with bootstrapped versions, even if they weren't deleted beforehand (e.g. on builder-hex0 with dummy unlink()) <Googulator>the only reason the renames were there is because originally they were move-if-changes <fossy>hmm, i think we might actually be in a good place to move toward 1.0, i just need to spend a few days doing a bit of a sweep through everything to make sure there's nothing critical i'm missing, but expect to see a 1.0 branch within hopefully a week or two (to allow us to still keep merging larger changes into master, and just do small updates/bugfixes in 1.0 branch) <fossy>i think i'd like to gather a bit of a "compatibility report" for bare metal, perhaps in the wiki section of github, so we don't waste time with "bare metal doesn't work on my X system" <Googulator>Although it would be great to have a way to actually do the "Only copy..." steps in a way that doesn't involve manually tracing build steps in the manifest <fossy>stikonas: yea README needs a bit of updating <fossy>Googulator: interactive mode addresses a fair few of them, but i'd like to add proper prompts for uninitialised variables in script-generator <fossy>well, interactive mode + the script-generator variable prompting <fossy>parts.rst needs a bit of updating too <Googulator>(remember to update the nss version number when we do the final nss update) <Googulator>stikonas: do you know if Guix has the same policy against pregens that LB has? <Googulator>(they might be interested in the Nyacc regen in that case) <stikonas>generally they run configure scripts, use bison files, etc... <stikonas>I don't think they are against improving stuff, but nobody works on that <stikonas>and also they mostly switched to mes/gash and replaced bash/coreutils, etc... <Googulator>It (or rather, Lisp) is "THE original MIT hacker language" after all :)