IRC channel logs
2023-01-17.log
back to list of logs
<doras>fossy: would you mind doing one bwrap bootstrap before merging the Python PR? <doras>To see that it works and that you get the same package hashes as the other bootstrap modes. <stikonas>though I'm also running Gentoo upgrade, so somewhat low on CPU resources... <doras>bwrap should be the quickest bootstrap mode, CPU-wise. <doras>Well, yeah. They should be similar. <stikonas>I did notice some strange things with bwrap mode, but didn't care to investigate <stikonas>early sed in bootstrap mode is quite slow on my system <stikonas>but a few seconds per sed command was not worth investigating <doras>bwrap is doing less stuff in RAM early on because we can't create tmpfs mounts. <stikonas>in any case, qemu mode is slowest both because it's slower to do each thing due to emulation <stikonas>but also because it builds more stuff (linux kernel) <doras>I mean, we do create a few tmpfs mounts as part of the bwrap invocation, but I vaguely remember that there is some gap there. <stikonas>anyway, it is a useful mode to test, I mostly run bwrap and qemu <stikonas>because if chroot fails, bwrap would also fail, but not necesserily vice versa <stikonas>and qemu run somewhat different stuff, part of it runs on a different kernel <stikonas>anyway, before we can merge that PR, at the very least we need to fix parts.rst <doras>bwrap is a bit more strict about file permissions and ownership, yes. <doras>I personally don't like the concept of a script running as root creating devices and executing a bunch of foreign build scripts with root privileges. <stikonas>but luckily live-bootstrap is not too needy here <stikonas>ideally building stuff shouldn't need root <stikonas>though sometimes I couldn't think of any alternative (not in live-bootstrap) <doras>We can probably limit the bwrap sandbox further if we wanted. <doras>I kept it pretty light-weight. <stikonas>i.e. creating rootfs images usually needs root privileges <stikonas>fossy: what do you think about expanding "Step 1" in parts.rst? <stikonas>fossy: also probably worth opening bug report for Python 3.11 regarding make regen-all <stikonas>well, write a bit about stage0-posix steps <stikonas>I tried to write a bit like that in stage0-uefi, but it's obviously slightly different text <fossy>yeah, that looks reasonably good to me <stikonas>anyway, once you fix parts.rst, I guess we can merge it <stikonas>I still need to look a bit more at pregen files <fossy>i just need to double check python 2.5 is fixed. it had a QEMU/chroot incompatibility <stikonas>though today I only have time for brwap mode <fossy>no real rush, i'd prefer to take the time to test it even if it takes a couple day <stikonas>by the way how much time does python build add? <stikonas>but we have quite a few builds of python <fossy>most of it is the newer python versions, they are massive <fossy>gah, py2.5 is not fixed. so checksum for that will still be wrong <fossy>wait, did i even push checksums <fossy>i'll push them when this build finishes <fossy>because hopefully, they should be the correct ones