IRC channel logs

<gm6k>I'm looking for pointers on the current status of Scala bootstrappability.

<gm6k>I'm possibly interested in doing some work on it, and https://bootstrappable.org/projects/jvm-languages.html suggests there is some preliminary work somewhere.

<stikonas>hmm, at least I'm not very up to date on where scala bootstrap currently is, maybe somebody else woud know

<stikonas>that page is also a bit out of date I guess...

<sh4>where's the maintained GCC 4.7.4 fork mentioned on the site?

<Googulator>It's 4.6.4 actually - there's a PR ongoing to have that corrected and the fork linked

<Googulator> https://github.com/ekaitz-zarraga/gcc

<sh4>why 4.6 ? 4.7.4 is the last C based one

<sh4>btw ekaitz_ you might be interested into the 4.7.4 patches done by mcf, the author of cproc https://github.com/michaelforney/gcc/commits/cproc-fixes/

<sh4>basically makes gcc source compatible with iso C99

<ekaitz_>sh4: I am! i have a few of those myself

<ekaitz_>ucontext_t is a real classic

<sh4>yah, fixed it several times myself

<ekaitz_>sh4: about why 4.6.4 it was because it had a modular distribution

<ekaitz_>and because somebody suggested that to me

<sh4>you mean the split between -core and other tarballs ?

<ekaitz_>i think that's what that somebody meant

<sh4>i dont really see an advantage to that split

<sh4>you want e.g. c++ support anyway

<ekaitz_>maybe for guix it is

<ekaitz_>the somebody is janneke maybe he has more info

<sh4>btw it might even be used for bootstrapping java in place of jikes

<sh4>i used to compile classpath with it

<ekaitz_>i didn't really pay a lot of attention to that part, I was quite overwhelmed by the work at the time

<ekaitz_>ACTION is now overwhelmed by work as a service

<sh4>what you mean by that?

<sh4>"work as a service"

<ekaitz_>i'm a freelance and basically all i do is too difficult until it isnt

<ekaitz_>and I'm always in things that are superhard for me, idk why

<ekaitz_>maybe because I run away from easy things, but I'm always in the verge of giving up

<ekaitz_>at the time I forked gcc I wasn't that used to it

<ekaitz_>sh4: i meant "overwhelmed (by work) as a service"

<sh4>well, a maintained gcc 4.7.4 fork is certainly an important cornerstone for the entire bootstrap chain

<sh4>optimally with a build system that sucks less, but that will be a lot effort...

<sh4>one could even try to backport some functionality (like C++ features) from newer versions that didnt use C++ features heavily

<ekaitz_>for me, it was very very hard at the time and most of the time i didn't know what I was doing

<ekaitz_>i'm still surprised it works

<sh4>i see

<ekaitz_>"works"

<sh4>so maybe it'd be better to start anew from the michaelforney fork ?

<sh4>as the author of a C compiler, he certainly knew what he was doing ;)

<ekaitz_>sh4: do you *really* believe that?

<ekaitz_>the worst part of gcc is not the code really

<sh4>yeah i know the guy well, worked with him on some projects

<ekaitz_>but the build system and where to put things

<sh4>right

<sh4>it's a major pain

<ekaitz_>most of the backend is defined in the .md files

<stikonas>4.6 or 4.7.4 doesn't make much difference...

<ekaitz_>you can probably take mine and rebase it on top of his

<stikonas>it wouldn't unlock any new gcc versions that you can build with 4.7 but not 4.6

<stikonas>and no new kernel versions

<ekaitz_>carefully, of course

<sh4>4.7 adds armhf and sh4 support

<sh4>armhf is a must have for any halfways recent 32 bit arm systems

<stikonas>well, for those arches you could still build 4.7

<stikonas>ekaitz only added riscv patches for 4.6

<stikonas>so your bootstrap path will diverge slighltly

<sh4>i took the info from the chart here https://github.com/GregorR/musl-cross which i once made

<stikonas>not that we have any bootstrap path for sh4 or arm

<sh4>ah, i thought this was a community effort, not guix-only

<stikonas>riscv gcc backport was mostly one person effort (ekaitz)

<stikonas>there isn't anything else in that fork

<stikonas>so hard to call community :)

<sh4>that patch could probably easily ported to 4.7

<ekaitz_>surely, yes

<ekaitz_>there are a few weird ones you sould skip

<stikonas>yeah, it's not that big

<ekaitz_>but the md and such you could add easily to 4.7

<stikonas>but it does need a bit of work, doesn't apply cleanly

<ekaitz_>sh4: if i had to do it, i would start copying the md files and then play with the buildsystem until it works

<ekaitz_>i wouldn't try to copy all the patches I made

<sh4>right

<sh4>your musl patches seem to be the old ones from musl-cross, not the new ones from musl-cross-make

<sh4>btw 4.2.4 might also be an interesting target, it was the last version that didn't need gmp, mpc, mpfr

<ekaitz_>those are some review we made later, not when I ported gcc but when we did the riscv bootstrap

IRC channel logs

2025-11-26.log