***ChanServ sets mode: -o janneke
<clemens3>stikonas: hi, i installed gentoo to try out your gentoo-bootstrap repo, but I have no clue about emerge/ebuild, how to use it, and it seems to be intended to be used by gentoo oldtimers.. <clemens3>any idea how to use the git repo? interested right now in the dev/java/jikes and on.. <stikonas>well, this one is portage overlay for Gentoo <stikonas>but Gentoo ebuilds are basically bash scripts, so you can kind of extract steps from them <clemens3>hmm, so how to i use portage to execute them.. and do they include all the jikes and classpath data, or it gets downloaded and gpg checked then? <stikonas>I'll be away for a few hours though, so you can try and if something is not working, I'll try to answer later <stikonas>it's already in the list of repos it knows about <stikonas>right now I'm still tweaking openjdk12 build <clemens3>i got manually stuck with classpath 0.99 dev version ***ChanServ sets mode: +o rekado_
<rekado_>clemens3: feel free to ask about the bootstrap. I implemented it in Guix a few years back. <clemens3>was stuck on guix with the nosubstitutes flag, so then went to gentoo, rebuilding now aka building llvm etc.. in case i need x for java support <clemens3>so will report later on that front, learning the emerge, repo use flags basics of gentoo atm <stikonas>well, on guix I'm getting stuck with openjdk-16 with nobusbstitutes <stikonas>clemens3: I've just pushed GCC 10 fixes for openjdk-12, so it should now work all the way till openjdk-17 <clemens3>i guess your java wants some gtk denpendency? anyway, X just finished <stikonas>if it disappears from gentoo, I need to move it to my overlay <stikonas>clemens3: let me ty to swap dependencies <stikonas>so maybe if I swap them, portage will work better <clemens3>stikonas: i did jikes, gnuclasspath 0.93.. i can go manually in small steps, if you tell me which order <stikonas>or at least it should be able to do directly icedtea:8 <stikonas>layman and eselect-repository are two similar products <clemens3>yeah, let me know when you did your change(s) <stikonas>let me know if it works, because I pushed untested change <stikonas>I still need to fix musl bootstrap in that overlay <clemens3>seems i need to disable one or more use falgs <stikonas>so then we can confirm whether the problem is with icedtea-8 ebuild or somewhere earlier <clemens3>i added --autounmask-backtrack=y to get full error message, then maybe more.. <clemens3>can it be i need use flags introspection, alsa, cups, gtk, jbootstrap? <stikonas>well, adding use flags can only complicate things <clemens3>maybe more, but i am in qemu, with less i dont see the colors <clemens3>without adding i couldnt install xorgs stuff.. i gues i need them, not? do you have minimal use flags yourself? <stikonas>maybe alsa, cups or gtk or something similar pulls in Xort stuff <stikonas>well, jbootstrap does not add any new dependencies, it just rebuilds icedtea with itself <clemens3>na, then i just do and then we see later.. will take a while to update @world, but what <clemens3>i need exactly one of (python_single_target_python3_8 python_single_target_python3_9 or 3_10 <stikonas>ok, let me try to construct smarter USE string <clemens3>gentoo and guix not exactly nooby friendly :) <stikonas>and gentoo has some circular dependencies that you can unbreak with fewer USE flags <stikonas>probably something needs python in earlier dependencies <stikonas>clemens3: oh, I know what might be wrong <stikonas>I thought maybe it's installing icedtea:8 not from overlay <stikonas>oh, I think I can reproduce the problem now <stikonas>clemens3: there are two places where icedtea8 and icedtea-7 has to be swapped in dependencies <stikonas>these last few weeks I was doing quite a lot of changes to unbreak bootstrap with GCC 10/11 <stikonas>and I don't bootstrap every time from zero... <clemens3>i guess not many people test the bootstrap thingy <stikonas>oh, that's probably because fastjar is masked in my system <stikonas>and see if it works without any USE= string <stikonas>if not, I can try to turn off some use flags <clemens3>dont know if same, but coimplains about use flags <clemens3>i create one.. need scp over.. waita second <stikonas>(ant-core blocker in my paste is due to ant-core from my overlay and from the distro not being coinstallable. But that's fine for bootstrap since upgrades from my to distro version work fine) <stikonas[m]>Oh, that's coming from just gentoo tree, not my overlay <clemens3>re, the use flags? ok, i will do, and resync, will take bit.. <stikonas>well, here I have quite a lot of use flags enabled globally <stikonas>so I think if you enabled postscript cups X and text globally it would work <stikonas>or if you build icedtea with USE="-cups" it might also work <stikonas>but yes, USE flags are a bit not user friendly <stikonas>although they allow great deal of customization <stikonas>because you can see it tries to use packages from my overlay (jamvm, classpath, etc) <clemens3>wait, postscript and text i havent yet in the set.. wait <stikonas>probably none of this is necessary for bootstrap <stikonas>i.e. printing support is definitely not needed to rebuild <stikonas>yeah, these ebuilds are not the most minimal... <stikonas>they are mostly taken from older portage tree checkouts and modified <clemens3>ok.. my purpose is what you and the guix guy did for gentoo and guix to apply it later to LFS <clemens3>so first is to be able to build it somewhere first <stikonas>yeah, well, with two examples it should be easier <stikonas>one is in scheme, the other is kind-of bash <clemens3>and debug process steps and apply to the other system hopefully <clemens3>stuck somewhere with javah and glasspath 0.99 dev <stikonas>well, on Gentoo I rebuild jamvm a few times <stikonas>each time pointing it out to different compiler / class library <clemens3>so comeing from LFS i am used to be a bit patient at least <stikonas>guix does some other magic with some jamvm wrappers <stikonas>yeah, it's the 2 one that I rebuil 3 times <clemens3>but there it seems also not so tested out the nosubstitute plus bootstrap <stikonas>each time installing into a different directory <stikonas>well, for me latest guix went just fine till openjdk:15 <stikonas>but 16 fails to build with some strange can't fork error <stikonas>somehow one would think I'm running out of process limit <clemens3>i installed guix package manager on lfs but then the installation with nosubstitutes hang later on <stikonas>well, but it's probably still the dependencies... <clemens3>yes, without resync, but either postscript or text was missing <clemens3>i can later give you my make.conf in case someone else comes along from an barebones gentoo install <stikonas>well, yeah, but you know it's doable to bootstrap ir <clemens3>basically all distros now depend on rust <stikonas>but at least mrustc is better maintained <clemens3>used pretty much in abiword and gnumeric <clemens3>and some personal stuff.. never tried, but open source search engine, distributed, based on java <clemens3>maybe let it sink.. i think i pack it for today <clemens3>if i can send more logs or env stuff, or reproduce something.. let me know.. 10 pm here now, gonna have to wrap up some other stuff, <clemens3>but definitely some bigger chunks moved forward for me.. <clemens3>ok, enjoy, and thanks for the help so far.. <stikonas>probably brok things in one of the commits today <stikonas>but need to build all the way up to icedtea-8 <bauen1>my hex0 baremetal clone is now working from spi flash, and is "only" 304 bytes, now the question is where i go from here, probably a hex0 with support for comments and some size reduction for boot0 <bauen1>i suppose it does support comments, if you avoid upper case letters and numbers <stikonas>well, adding proper hex0 comments as in stage0 is just a few lines of assembly <stikonas>I think the bigger problem is input/output <clemens3>hey stikonas cool, will update tomorrow! <stikonas>or if you ignore that and just assume spi are programmed externally, then I guess follow stage0 path to hex1 and hex2 <stikonas>clemens3: ok, I'm now building icedtea:7 <stikonas>well, I should't push untested patches... <bauen1>stikonas: yes, only boot0 (implementing a very rudamentary hex0 and uart0 driver) will be hand toggled into SPI Flash <bauen1>UART is kind of cheating, but apart from that I want to pursue slightly different goals than live-bootstrap and supply the tools necessary to audit code, then build and use it <bauen1>so e.g. an sha2 implementation will be one of the first things to be done <stikonas>but so is live-bootsrap, even more so... <bauen1>stikonas: i'm not sure what alternatives there are, it might be possible to write an hdmi driver for this chip and create a framebuffer that way, and a usb + usb-hid driver <bauen1>stikonas: and also some way to store data, even if little ; first by writing an SPI driver and writing to SPI flash and then I'll look into driving the SD-Card maybe <bauen1>stikonas: i mean, you don't need fancy graphics, you only need a TTY <bauen1>but maybe building a TTY from logic chips and generating a VGA signal would be easier :D <bauen1>yeah, i'm looking up how many pixels that has <bauen1>like you want at least 80 columns and a minimum of 25 lines <stikonas>80 columns with 25 lines is proper terminal... <bauen1>true, but if you want to review any amount of code you want an acceptable terminal <stikonas>well, you can build acceptable terminal once you reach C compiler... <bauen1>a different approach would be to only do the implementation up to sha2 on a "trusted IO device", and then assume that if the hashes are good then the code must be good <bauen1>that's not how you'd want to do it the first time around, but certainly if you're iterating / developing <bauen1>only hand audit up to your sha2 implementation and after that let the kaem equivialent take over <doras>NickServ wanted me to identify. Matrix didn't do that automatically for some reason. <doras>And yes, it's in bwrap mode. <doras>But not in the "normal" bwrap mode that runs with a fake root. <doras>It happens only in the one where the normal user is doing the build. <doras>I ran into a few permission-related issues that I solved. <doras>But this Perl thing, I can't figure out. <stikonas>why are we not allowing not identified posters here <stikonas>you have op, so if we want unidentified people, we can enable it <stikonas>doras: if you can, maybe try to compare files between good and bad builds <doras>I'll need to remove the sysa cleanup. <doras>stikonas: I just verified that faking root solves it. <doras>The nonsense header I pasted above about #running_under_some_shell is basically gone when things work <doras>Instead all I see in its place is: <doras>I'll paste the good and bad file quickly. <doras>You can see that the beginning of the file is a broken mess in the bad case. <doras>I think "help2man.PL" file generates this. <stikonas>yes, it looks like help2man.PL generates bad code <stikonas>now the question is, why this workaround doesn't work without root <stikonas>or possibly the better question is why things go wrong in the first place <doras>You think the workaround is broken then? Interesting. <doras>Let me confirm it for a moment. <stikonas>yeah, I'm now starting to recall something like that <doras>I know why it's broken if it is, let's say