17:09 <zerdox> chadmed: so.. make LLVM=1 rustavailable itself gives me syntax error (scripts/rust_is_available.sh line 21). i did all requirements except older rustc, but still don't get hidden option in menuconfig

17:11 <MichaelLong> that part is a lot of work

17:11 <zerdox> sorry about me being this but... i can't even google it

17:12 <MichaelLong> zerdox, no problem I've another error compiling the kernel with the rust bindings. don't have it at hand right now.

17:12 <zerdox> exactly i get line 21: 100000 * + 100 * + : syntax error: operand expected

17:18 <zerdox> it's get_canonical_version which uses $1-3 inside which are empty for some reason

17:26 <zerdox> ok so for now i will just wait for your ping...

17:26 <zerdox> and thank you

17:31 <MichaelLong> zerdox, so your're also on gentoo or another distribution?

17:59 <ChaosPrincess> zerdox: make sure rustc is the version it wants and is in PATH, also bindgen is the correct version

18:00 <ChaosPrincess> whoever wrote the makefile that way is a bad person, :P, having a bunch of drivers randomly vanish because you dared to run make menuconfig w/o rust in path is terrible

18:08 <MichaelLong> in some kernel include file.

18:08 <MichaelLong> I used this https://codentium.com/building-a-linux-kernel-with-rust-support-on-gentoo/ (now older) article, but stopped as I have rust 1.66.1 (the kernel would like to have 1.66.0) and bindgen is available in portage in version 0.56.0 so I was save. I was able to overcome the "make LLVM=1 rustavailable" stage and can select rust in the kernel eventually but compiling the kernel failed very very early complaingin rust having some "indent" error

18:10 <MichaelLong> (don't have the exact failure message right now as I shut down the laptop for today)

18:13 <zerdox> MichaelLong: i am on gentoo. i would not like to compile kernel but it's maybe required for me to graphics work correctly

18:13 <MichaelLong> zerdox, I fear it is :)

18:14 <zerdox> ChaosPrincess ty. i guess i will remove emerged rust and will use rustup because i already cannot downgrade rustc

18:14 <ChaosPrincess> im using emerged rust for stuff that needs it to build and a separate rust for kernel (that rust is not in PATH unless im compiling kernels)

18:15 <ChaosPrincess> this is all brittle and janky, but technically works

18:15 <MichaelLong> but why?

18:16 <MichaelLong> was I was also almost going that way but looking at packages.gentoo.org virtual/rust is already stable at 1.66.1. the ashai-kernel I use (based on 6.2.0) wants 1.66.0

20:52 <chadmed> MichaelLong: for a while the kernel rust was unstable and new versions would throw errors every week so the recommendation was to have rustup override your system toolchain

20:52 <chadmed> but ive been using the packaged ~arm64 rustc for ages now and it works fine

21:07 <MichaelLong> chadmed, a thx for the clarification. that stuff is totally new for me, just trying to get it working.

22:42 <chadmed_> MichaelLong: we're all learning, dont sweat it :P

22:43 <chadmed_> ive been racking my brain trying to think of how we handle this for packaging rusty sources upstream and i think the easiest way is make a rust ebuild for bindgen

22:43 <chadmed_> but theres already one for cbindgen (different) because firefox uses it

22:43 <chadmed_> so its a mess

22:54 <jannau> it's also a mess because rustc/bindgen are neither build nor normal dependencies

22:55 <jannau> normal dependencies are probably the better fit but would require slotted rustc/bindgen

22:57 <chadmed_> rustc is slotted now i think?

22:58 <chadmed_> it doesnt really matter for us (our systems are very fast) but bootstrapping even one version of rust is a tall order for like 90% of systems

22:59 <chadmed_> i think i put an RDEPEND on a specific version of rust behind the

22:59 <chadmed_> "rust" USE flag in our ebuild

23:00 <chadmed_> with the gpu coming down the pipeline ill start thinking a bit harder about it

23:00 <jannau> ah, rust is slotted but virtual/rust and rust-bin are not

23:00 <chadmed_> sam_- might have some ideas

23:01 <jannau> bindgen is hopefully less of a problem once the kernel is updated for allow/block-list