IRC channel logs

<stikonas>rickmasters: I fixed that initramfs issue

<stikonas>actually I would suggest not reverting run2.sh but for now move bash 5 to sysc and rebuild bash 2 in sysa

<stikonas>medium term we probably will be able to upgrade to bash 4 in sysa, so there will be less refactoring of run.sh vs run2.sh

<nektro>doras, woah congrats :)

<muurkha>doras: that's wonderful! congratulations!

<fossy>oh, sorry about #259. that was a silly mistake with git from me :\

<fossy>rickmasters, stikonas[m]: i'm happy with rebuilding bash 2 instead.. i'll remake my branch to do that

<fossy>oh

<fossy>stikonas already did it haha thanks :D

<fossy>oh but not the actual bash 5 removal

<fossy>ok got it ill do that bit now

<stikonas[m]>fossy: no, I just patched automake

<stikonas[m]>Cherry picked your commit and fixed linter

<stikonas[m]>We still need to add bash 5 to sysc and rebuild bash 2 in sysa where bash 5 currently is

<stikonas[m]>OK, merged

<stikonas[m]>rickmasters: can you try tip of live-bootstrap now?

<gforce_de1977>current master fails:

<gforce_de1977> https://github.com/fosslinux/live-bootstrap/issues/261

<gforce_de1977>stikonas: question: are we downloading files during bootstrapping? is'nt it always prepared on sysa before qemu starts?

<gforce_de1977>stikonas: it other words - we need network access?

<fossy>gforce_de1977: that is the default to download distfiles, yes

<fossy>you can use --external-sources

<fossy>which uses the old behaviour of copying distfiles in externally

<fossy>this is not final behaviour by the way

<fossy>possibly subject to change at a later date

<gforce_de1977>fossy: thanks for making that clear - i was'nt aware of that

<gforce_de1977>wouldn't it be great to print the real checksum when in fails, e.g. in sysa/helpers.sh => src_checksum()

<stikonas>gforce_de1977: we have --update-checksum mode that ignores checksums

<stikonas>I don't think we can print real checksum when it fails

<stikonas>traps are broken in bash 2

<stikonas>fossy: I've now got strip: unable to copy file '/tmp/destdir/usr/lib/perl5/5.32.1/i386-linux/CORE/libperl.a'; reason: Permission denied

<stikonas>I'll retest but it's unlikely to be one-off

<AwesomeAdam54321>The stage0 bootstrap path achieves binary-source correspondence in bootstrapping by making the source correspond to the binary

<AwesomeAdam54321>Is it feasible for an alternate bootstrap path to achieve binary-source correspondence by making the binary correspond to the source?

<AwesomeAdam54321>For example, Lisp Machines that could directly run Lisp code

<gforce_de1977>AwesomeAdam54321: maybe not for all (early) steps, but may i ask: which step would a lisp-machine run at first? consider: https://github.com/fosslinux/live-bootstrap/blob/master/parts.rst - which parts can be thrown away/subsituted by a lisp machine?

<stikonas>well, you don't need early steps for those machines that can run source

<stikonas>though I guess any such hardware would be quite inefficient

<stikonas>which might be the reason why nobody is making it today

<stikonas>it's basically just moving compiler to hardware

<AwesomeAdam54321>gforce_de1977: Steps 1-13 can be skipped with a lisp machine, with mescc being the first step

<gforce_de1977>AwesomeAdam54321 - i theory there should be no differences with source and binary for "traditional" and "lisp-machine" bootstrap.

<gforce_de1977>nice experiment anyway, but IMHO there a little bit of fun (the early steps) missing 8-)

<gforce_de1977>AwesomeAdam54321 - do you HAVE a lisp machine?

<AwesomeAdam54321>gforce_de1977: No, it was just something I was thinking about

<gforce_de1977> https://github.com/ghosthamlet/awesome-lisp-machine

<gforce_de1977>but it must be a scheme machine, right?

<AwesomeAdam54321>definitely

<stikonas[m]>AwesomeAdam54321: is mescc even useful for lisp machine?

<stikonas[m]>That's C compiler

<rickmasters>fossy: Unfortunately, the bash 2 rebuild is failing on Fiwix.

<rickmasters>fossy: I get a couple page faults running configure and then compilation fails with a bison general protection fault.

<AwesomeAdam54321>stikonas[m]: It's useful if you want to build packages written in C

<stikonas[m]>rickmasters: hmm, that's a bit unfortunate :(

<AwesomeAdam54321>But mescc doesn't target the lisp machine so it is actually useless

<stikonas[m]>How does it work on fiwix is then...

<stikonas[m]>rickmasters: should we remove bash 2 rebuild then?

<stikonas[m]>I meant how does it work on fiwix os

<rickmasters>stikonas: I don't know. It may use a different bison.

<stikonas[m]>Oh maybe they don't run bison at all

<stikonas[m]>Still, we used bison quite a bit already

<stikonas[m]>rickmasters: can you comment out bash rebuild step?

<stikonas[m]>So that we can see if things after it work

<rickmasters>stikonas: sure

<rickmasters>stikonas: will the `exec env -i bash run2.sh` command still work at the end of run.sh?

<stikonas[m]>It should

<stikonas[m]>Even if we remove bash rebuild, we could keep that exec as future investigation placeholder

<avih>oriansj: so what do we do about strict hex0? do you want to drop it? maybe take only the revised definition of the current hex0? ("loose")

<avih>(i want to fix the shell script)

<rickmasters>stikonas: it works fine with bash 2 rebuild commented out.

<stikonas>all the way to Linux kernel?

<stikonas>ok, let's disable it for now

<stikonas>there is probably a way to workaround it but let's get fiwix work done first

<stikonas>(i.e. yacc that we built early in the bootstrap did manage to build bash's parser

<rickmasters>stikonas: yes it builds linux

<rickmasters>stikonas: I'm willing to look into the bison problem down the road a bit.

<stikonas>let's merge it first...

<rickmasters>stikonas: well, I've written the code to boot Fiwix off disk. Just need to make it work now.

<Guest34>janneke yeah, I mailed for subscription. As I'm a newbie can you tell where should I start from

<Arghadeep_Saha>Same here can anyone tell me where to start from

<stikonas>if you want to work on mes, janneke or ekaitz might tell you

<stikonas>though mes is not the simplest stuff to begin with

<Arghadeep_Saha>so could you suggest some simpler task to begin with

<stikonas>maybe just start by building mes

<stikonas>see if you can run it

<stikonas>(there are various ways to build it, you could build it using gcc, but if you want more bootstrappable type experience, you could first try to build M2-Planet and then use it to build mes

<stikonas>we do have everything automated (e.g. this repo https://github.com/fosslinux/live-bootstrap/blob/master/parts.rst builds mes), but you could try running some of the steps manually

<Guest34>stikonas: thanks for your reply. Will try to implement it. Since I still don't know much about it. Thus, will take some time to study it and get some understanding.

<Guest34>will documentation be a good start?

<stikonas>yes, reading README files is always a good start

<stikonas>but again, if you want to look at mes, wait for janneke for more detailed advice

<Arghadeep_Saha>yes sure, thank you for your advice.

<stikonas>readme file of mes has some further links too https://git.savannah.gnu.org/cgit/mes.git/tree/

<janneke>hmm, that was fast

<ekaitz>stikonas: I love the way you guys think I know what I'm doing :D

IRC channel logs

2023-03-11.log