<fossy>stikonas: so, sysa kexecs sysb, which is an initramfs, which purely switch_root's to sysc <fossy>we do this because we cannot assume sysa's kernel has hard disk support <stikonas>fossy: do you want me to wait with guile until you merge your changes? <fossy>preferably :) then we don't need to worry about the ram issue <stikonas>I think I've fixed the problems, but I'm concious that you would be playing catch up <stikonas>can't test whether it's reproducible on 32-bits <stikonas>btw, I have a small PR to add trap to bash <stikonas>it might break current CI stop detection thoughbut current CI is broken anyway, we should probably remove run step <Hagfish>switching to sysc via sysb is so elegant ***cadmium.libera.chat sets mode: +o ChanServ
<fossy>award for easiest-to-remove pregened files thus far goes to the Linux kernel <fossy>find . -name "*_shipped" -delete <stikonas>fossy: is it something reasonably newish? <stikonas>there are actually more files that are automatically generated, but probably out of scope for live-bootstrap... Especially in drivers <stikonas>at least in modern kernel there are hundreds of autogenerated lines of code that we have no means of rebuilding <stikonas>e.g. linux/drivers/gpu/drm/amd/include/asic_reg has 164MB of autogenerated defines... <stikonas>oriansj: well, linux libre removes binary blobs <stikonas>there is nothing we can do about other generated files, I think we just have to accept them <stikonas>but yes, that's why I asked fossy which kernel is used in his work <clemens3>yeah, quality is subjective, but bloat, obfuscation, complexity, intended or not, have the same result as binaries and blobs.. <clemens3>nobody can follow what is going on.. so simplicity/understandability etc. needs to be a goal.. not just source <integfred>Hi I am author of myixos.org ... I also wrote something else, which is why I'm here regarding Maven Central <integfred>I spoke to someone called Julian a long time back on IRC <Hagfish>i don't know what the situation is with Maven Central and reproducibility, but that's an interesting question <integfred>So these scripts and program I wrote for maven central, attempts to 1) Recursively download sources given an artifact 2) Reconstruct the Maven software by iterating over the java classes using javac <integfred>I also tried to do some auto-patching from disassembled bytecode <Hagfish>auto-patching? yeah, that sounds nasty <integfred>I will post the scripts probably next month, I'll come back here when I've uploaded them to git <integfred>They only yield about 10% of the classes. I didn't actually manage to reconstruct a whole piece of software. <integfred>Complete ball ache trying to upload to github, but never mind. <Hagfish>i don't know if that should be mentioned on the bootstrapping wiki or something <Hagfish>it might be the sort of thing that someone else might be desperate to contribute to if they saw it, as it might scratch an itch <fossy>stikonas: hm. I did forget about thay <stikonas>well, ideally we at least get rid of binary blobs... <stikonas>driver shipped pregenerated headers with no original source are probably beyond us here... <stikonas>I think linux-libre is available since 2.6.27 <fossy>quick looks at history seems that gnewsense did deblobbong around this time of kernel <stikonas>there is 2.6.21 release of linux-libre in old/gen1 folder <stikonas>I guess that's when they took over from gnewsense <stikonas>2.6.21 I guess is a bout a year or so newer