IRC channel logs
2024-02-18.log
back to list of logs
<rickmasters>Meanwhile, I ran into a checksum error this morning building libtool-2.2.4_0.tar.bz2. <rickmasters>Unfortunately, I had to run 15 more builds until I could reproduce it. <Googulator>rickmasters: "AUTOMAKE=automake-1.10 ACLOCAL=aclocal-1.10 AUTOM4TE=autom4te-2.61 AUTOCONF=autoconf-2.61 AUTOHEADER=autoheader-2.61 AUTORECONF=autoreconf-2.61 ./bootstrap" <Googulator>sounds like that's the same issue I've seen before <Googulator>otherwise it will do weird non-reproducible things <Googulator>btw, that particular curl failure happened in bwrap mode, but I do remember curl failing on bare metal too, dropping to the trap shell <Googulator>as for the savannah tarballs issue - IMO we should download a set of tarballs once, and then rehost those on archive.org <Googulator>this way, we can use xz tarballs, which are notoriously unreliable on savannah itself <rickmasters>Googulator: Can you elaborate on the autoreconf problem? The line you quoted is running bootstrap, which runs autoreconf --force <Googulator>and because we patched configure.ac in that package, that patch would not get applied to the real configure script when autoreconf decided there was no need to regenerate it <Googulator>but if it's already called with --force, that's different <Googulator>although there could well be something else in the bootstrap script that needs a --force option to make sure it doesn't just decide "it's up to date, I'm not gonna touch it" <rickmasters>Got it. It seems to have --force so I'm suspecting a timestamp / make related issue. Not sure yet. <rickmasters>This happened on kernel bootstrap and Fiwix uses ext2 which has coarse timestamps. We've seen that with other packages before. <rickmasters>But it's too early to tell. I'm still trying to reproduce it better. <rickmasters>With regards to the bash trap, I don't seem to be getting a prompt. I was testing bwrap, chroot, qemu/linux, qemu/kernel bootstrap. <rickmast_>They all had a curl fail at the same time, no prompt though. <Exaga>thanks for the heads up pabs3 <Exaga>i will certainly look into that <pabs3>I think their channel is on OFTC, but I forget the name <stikonas>mid-kid: thanks, I've bookmarked it now. Nice work! <fossy>mid-kid: hm, what if you built up a no-multilib system and then used that to build a multilib system? (i have veyr little knowledge of gentoo multilib, so i might be spouting stupidity :D) <fossy>either way, this is very impressive!! <fossy>rickmasters: i will give a go at the bare metal issue now :) <fossy>Googulator: regarding the archive.org idea, yes i am OK with that, on the condition that the archive.org tarballs consistently produce the exact same directory structure as upstream <fossy>Googulator: I've gone through the kernel config and made a few notes on the PR. I don't feel particularly strongly about any of them, and the removals are just to save space, but let me know what you think for each <fossy>rickmasters: unfortunately that commit didn't fix it <stikonas>argh, another stupid error on my openjdk gentoo overlay that I'm now retesting... <stikonas>/var/tmp/portage/dev-java/icedtea-7.2.6.9/work/icedtea-2.6.9/bootstrap/jdk1.6.0/bin/java -XX:-PrintVMOptions -XX:+UnlockDiagnosticVMOptions -XX:-LogVMOutput -Xmx512m -Xms512m -XX:PermSize=32m -XX:MaxPermSize=160m -jar /var/tmp/portage/dev-java/icedtea-7.2.6.9/work/icedtea-2.6.9/openjdk.build-boot/btjars/generatecurrencydata.jar -o /var/tmp/portage/dev-java/icedtea-7.2.6.9/work/icedtea-2.6.9/openjdk.build-boot/lib/currency. <stikonas>data.temp < ./openjdk/jdk/src/share/classes/java/util/CurrencyData.properties <stikonas>Error: time is more than 10 years from present: 1388527200000 <sam_>building java is so.. challenging <sam_>mostly because of stupid stuff like this, not even anything reasonable <sam_>did you see the recent salt thing btw? <sam_>it stopped working *outside* of tests, too <stikonas>anyway, that 10 year thing is patchable of course <stikonas>hmm, I've never even heard of salt before <oriansj>well in September 2020, VMware acquired SaltStack; so level of surprise == 0; <oriansj>I've used it in a couple environments and once Nix came into existing, that whole category of tools lost their purpose. <Googulator>fossy: nice catch for BIGSMP - I was wondering why I can't set the max core count higher than 8, this is probably it <Googulator>as for the others: "Pentium Pro" is the trade name of 686 <Googulator>initramfs compression: the xz, lzo, etc. libraries still get included in the kernel for other purposes, so the space saving is minimal; on the flop side, keeping support here means we can in the future easily save RAM in Fiwix by decreasing the kexec buffer size and using higher compression for the initramfs <Googulator>CONFIG_DRM_I915 (and other DRM drivers) isn't just for X; DRM drivers are the recommended way of getting a high resolution framebuffer on modern kernels, the plain-FB drivers are deprecated-ish and can even cause problems. In fact, I plan on switching to nouveau/radeon/amdgpu and disabling their FB counterparts, because nvidiafb turned out to be <Googulator>the cause of the green screen freeze bug on newer GeForce cards. (And yes, this functionality works without any firmware loaded; firmware is only required for 3D acceleration.) <Googulator>KVM, unfortunately, doesn't compile with our current toolchain; I tried to patch it out, with no luck. It causes "undefined reference to `__compiletime_assert___COUNTER__'" errors at the final link, and unlike drm_edid, this doesn't seem to be patchable just by removing BUILD_BUG_ONs. <Googulator>As for mouse support, I don't think we need the savings from that right now, if it's even significant. I'd leave that for a future size reduction opportunity, if ever needed, especially since we already had mice enabled in 4.9.10. <Googulator>these are the files we currently pull from Savannah or GitHub <stikonas>and what do we need to do as we switch to different versions? <Googulator>Download once from the forge site, checksum, upload to archive.org, and link it from there in sources - checksums are then guaranteed not to change, unlike when downloading directly from GH or Savannah <matrix_bridge><cosinusoidally> Googulator: I find archive.org performace pretty variable. Some were ok, but others were < 500KB/s. Maybe it depends on where you are in the world (I'm in the UK) <Googulator>cosinusoidally: Savannah is far slower than that, if it even works <Googulator>GitHub is somewhat controversial as a host, because it's owned by MS <stikonas>Googulator: strange, I've got some build error on the latest ommit <stikonas>this was with ./rootfs.py --build-kernels --bwrap <Googulator>new version of unxz in stage0-posix/mescc-tools-extra <fossy>stikonas, Googulator: i said earlier regarding archive.org <fossy>regarding the archive.org idea, yes i am OK with that, on the condition that the archive.org tarballs consistently produce the exact same directory structure as upstream <stikonas>well, archive structure is a one-off copy of upstream <Googulator>more precisely, the archive.org tarballs are binary identical to what upstream was offering for download at some specific time <Googulator>if upstream chooses to alter how it makes tarballs, the archive.org ones don't change unless we explicitly update them <fossy>well, yes, however if for example a new file or something was added to the git-generated tarballs, say with some metadata (would be very silly, but nevertheless), then i'd be saying we should update the archive.org tarballs <fossy>it just adds an extra layer of ability for verification <Googulator>IMO that should be a new set of uploaded files on archive.org, with new URLs <Googulator>and an explicit commit to LB to point to those URLs <fossy>sorry yes, that's what i meant <fossy>i think we are on the same page here <Googulator>if we overwrite the originals on archive.org with changed ones, breaking checksums, then we're only doing marginally better than GH/Savannah <Googulator>(marginally better, because at least it's a consistent checksum fail, not a random one with some HTTP 5xx errors and random timeouts interspersed) <Googulator>but it's better to go all out and actually keep the original files at the original URLs, with new versions getting new URLs <Googulator>it also better aligns with archive.org's goals of being, well, an archive