<stikonas>so for those who don't follow it too closely, live-bootstrap has now G++ 4.7.4 merged <stikonas>fossy: I guess I'll now wait for your PR, otherwise if I do more PRs, then you'll be forever playing catch up.. <oriansj>stikonas: then you might want to update the parts.rst to match. <oriansj>wait, sorry missed the "we build both C and C++ backends" <oriansj>So all that absolutely remains to bootstrap guix is guile right stikonas? <stikonas>oriansj: we have guile in a slightly older branch (before we decided to build guile first), I think I'll rebase it now (drop half of the commits that are already merged + update checksums) <stikonas>hopefully guix can bootstrap with guile 3.0 <stikonas>guile and g++ are a bit orthogonal if autogen is avoided <stikonas>I haven't tried building guix in live-bootstrap yet... <stikonas>it might be best to do a after.sh hook for live-bootstrap too <stikonas>so that people can add extra on top, e.g. guix or something else for other distros <stikonas>(similarly to after.kaem hook in stage0-posix) <oriansj>stikonas: it might be best to have civodul help us form a new guix seed from live-bootstrap. <oriansj>now it is on me to solve the kernel problem. <dongcarl>Hi, apologies I've been out of the loop for quite a while... <dongcarl>Wondering if oriansj's message in #guix means that live-bootstrap will integrate into Guix? ***ChanServ sets mode: +o janneke
<stikonas>dongcarl: well, by integrate I mean live-bootstrap can build all of the guix bootstrap binaries on i386 (including static guile) so we have an option of replacing bootstrap binaries with built in live-bootstrap. (I don't know how bootstrap binaries are built now) <stikonas>guix binary/daemon can probably be easily built on live-bootstrap too... ***ericonr- is now known as ericonr
***bandali_ is now known as bandali
<clemens3>I find it a bit ironic, here you are thinking about making your own hardware in order to avoid getting a backdoor.. but when I want to install the guix package installer, it says download this install script, and.. run it! <clemens3>at least maybe provide it with a signed gpg key or such.. <vagrantc>clemens3: things can always be improved at nearly any stage in the process of bootstrappability <vagrantc>there's always another chicken and another egg <vagrantc>which came first, the chicken, the egg, or the yoghurt? <clemens3>vagrantc: trying to bootstrap java and reusing the guix groundwork.. <vagrantc>clemens3: but you can fix later parts in the chain and later go to earlier parts in the chain ... it's not always an either-or scenario <vagrantc>fix what you can where you can, and work towards improving it <vagrantc>basically, absolutely nothing is absolutely bootstrappable at this point, but there are parts of the process that are, with given starting points <clemens3>yeah, no problem, it is what it is..:) but I find it all over the place then things just get downloaded and executed, be it docker images, what not, and at the same time whole research departments go down on the micro code and what not.. <vagrantc>yeah, the "status quo" is pretty awful :) ***lfam is now known as dstolfa1
***dstolfa1 is now known as lfam
***pgreco_ is now known as pgreco
<dongcarl>stikonas: Oh you mean there’s a path to replace ALL of the guix bootstrap binaries with live-bootstrap? <dongcarl>I’m guessing Guile was the hardest part that just got finished? <stikonas>dongcarl: yes, I think so. I haven't tested binaries that live-bootstrap builds with guix but I guess it should work (somebody more familiar with guix can tell me) <stikonas>at the moment there is also gcc/g++ 4.7.4 and binutils 2.14 but if necessary latest binutils (I think 2.36) just builds. As for gcc, next step I think can jump up to gcc 10.3.0 <stikonas>(mabe not 11, since gcc 11 uses more modern C++ features) <stikonas>guile is indeed the hardest out of all bootstrap binaries (others like mkdir, tar, xz, bash are much simpler) <Hagfish>gcc 10.3 was released this april. that's huge progress if it can be reached <stikonas>well, with gcc is not hard to jump to newer versions <stikonas>once you have gcc/g++ 4.7.4 it's not a big deal to go to newer versions <Hagfish>that does make me feel a bit better about gcc, thank you ***ChanServ sets mode: +o oriansj
<oriansj>dongcarl: technically we have a path to replacing *ALL* bootstrap binaries for *EVERYONE* as once one has Guile, they are able to run Guix. Once one has GCC, the rest of the world of software is only a guix build chain away. (With various unknown bootstrapping nightmares to kill along the way) <stikonas>even without Guix, it's not hard for everyone to bootstrap from what is in live-bootstrap. It's already full i386 toolchain with a lot of tools that people expect <stikonas>it's mostly a case of building libraries that you use <stikonas>dongcarl: which might be a bit more work for more complicated projects, e.g. bitcoin would need Qt. But something with less dependencies is very easy to build <stikonas>that said, Guix has an advantage of a lot of packaged stuff that is more maintained <stikonas>although, guix does not do no-pregen stuff in packages but it's just a matter of nobody did that, not because tools needed for regeneration are not bootstrapped <stikonas>well, maybe with the exception of autogen... we don't know how to build autogen with no pre-generated files <stikonas>and of course we still have kernel that is not bootstrapped <oriansj>stikonas: we will get to that. Just need to keep solving one problem at a time until all the work is done (Which will never happen but it'll be fun to see the excuses get more and more pathetic)