IRC channel logs
2024-01-11.log
back to list of logs
<rickmasters>OK reproduced checksum fail for python 3.1.5 using latest live-bootstrap. The python versions before that were OK. <Googulator>One way you can speed up testing Python issues is by commenting out guile and everything else until Python <Googulator>Guile & Autogen are AFAIK only needed for GCC 10+ <rickmasters>Thanks. Just passed python 3.1.5 with Fiwix update with success. <rickmasters>I'll continue testing to find the revision that python 3.1.5 first fails on. <Googulator>I've done a test myself to revert the Fiwix upgrade - Python checksums are still wrong, matching the ones with the new Fiwix. <fossy>jenntech23: i think you are very much in the wrong channel. <muurkha>well, perhaps they are in the right channel but sent that message to the wrong one, or perhaps there's some meaning I haven't deciphered <Googulator>rekado: yale-haskell is failing in exactly the same way in clisp 2.49, the last proper release from 2010 <Googulator>by any chance, do you remember what version you used in 2016? <stikonas>hopefully no pregenerated files there :) <fossy>cosiunsoidally: wow, this is surprisingly a fairly simple PR. I was expecting a lot more hacks to be required with mes libc! <stikonas>well, I guess a lot is hidden inside tcc-bootstrap-alt repo <fossy>what libc is tcc itself compiled against? <fossy>that's only headers though... not any actual code <matrix_bridge><cosinusoidally> yep woody headers (which are fairly big). It can use the in tree polyfill libc or glibc, it's just a case of linking the object code to the desired one. <matrix_bridge><cosinusoidally> There are lots of headers I should probably strip out. <matrix_bridge><cosinusoidally> Checksums are disabled for mes (since it's not built) and tcc-0.9.26 (since for some reason they don't match). By the time it gets to the live-bootstrap tcc-0.9.27 the checksums are the same again. <matrix_bridge><cosinusoidally> During the bootstrap I use the polyfill libc obviously. <fossy>is the polyfill libc in M2libc directory? <matrix_bridge><cosinusoidally> There are several polyfill libcs and they are a bit spread out/overlapping. I think my next step is probably a bit of writing documentation for the process. <matrix_bridge><cosinusoidally> I also need to cut them down and deduplicate. <matrix_bridge><cosinusoidally> libc_boot.c is the first elf version of libc used by the first elf version of tcc. <matrix_bridge><cosinusoidally> The elf version of tcc will also call across to elf_loader_support_cc_x86.c (and then to some of the stuff in M2libc for certain low level stuff/syscalls) <fossy>regarding the PR itself, looks fairly sensible to me. there will need to be checksums done though <fossy>it would also be nice if the tcc_bootstrap & mes paths produced the same tcc binary at the end. i feel like that should be possible - and could already be the case - but then again, maybe not <matrix_bridge><cosinusoidally> I'm guessing it's probably some missing define somewhere. the 0.9.27 binary is identical, just not the 0.9.26 one. live-bootstrap ci checks all the checksums by default? If so, all the hashes after 0.9.26 are correct. <rickmasters>cosinusoidally: regarding: cp artifacts/tcc_27_boot_static.exe /usr/bin/tcc-boot0 <rickmasters>where does artifacts/tcc_27_boot_static.exe come from? <fossy>so currently tcc_bootstrap_alt mode is not protected by checksums <rickmasters>oh nevermind, I haven't looked at the new alt sources <rickmasters>the builds steps for tcc_27_boot_static.exe must be in there (?) <stikonas>it might be that we need to do more rebuilds for them to produce same binary... <stikonas>right now mes path has the minimum number of rebuilds until checksum stabilizes <stikonas>which is just tcc-mes, tcc-boot0 tcc-boot1 and final tcc <stikonas>and also version of tcc 0.9.26 is a bit confusing, it's much closer to 0.9.27... <rickmasters>cosinusoidally: thanks. there is ... a lot more phases there than I was expecting. wow <matrix_bridge><cosinusoidally> fossy: ah I see, I didn't check the sha sums for tcc_bootstrap_alt in its pass1.kaem. I can add that in. Would that just be the sha256sum for tcc_27_boot_static.exe / /usr/bin/tcc-boot0 . <matrix_bridge>Just going to check if the sha256sums for the live-bootstrap version of "tcc-0.9.26" differ. <rickmasters>fossy: next problem with --kernel is that mkdir and mount of /proc before kexec were left out of the simplify overhaul. <rickmasters>fossy: can you remember if there was any reason to remove the /proc stuff? <rickmasters>fossy: I can provide a PR for that. Unfortunately, there are checksum issues with tar-1.34 later on. So --kernel still won't be working yet. <fossy>yes cosinusoidally, that is correct <fossy>rickmasters: oh, mount of /proc is done in populate_device_nodes. (probably hsould change the name of that script). anyways a improve: populate_device_nodes should be added between build: util-linux-2.19.1 and jump: linux <rickmasters>fossy: ok, is right before linux ok? I'll test it first... <fossy>right before linux should be ok