| 25 Dec 2024 |
 DavHau | Fixed a bunch of cross build issue with python packages and got home-assistant almost running, but there is one I'm having trouble with wrapping my head around. The python bcrypt library just builds its shared object for x86 instead of riscv
Any ideas why?
> nix-build -A pkgsCross.riscv64.python3Packages.bcrypt
/nix/store/xfc8xssvq2vwcn6126ml3m6fkdd1gzw0-python3.12-bcrypt-4.2.0-riscv64-unknown-linux-gnu
> file /nix/store/xfc8xssvq2vwcn6126ml3m6fkdd1gzw0-python3.12-bcrypt-4.2.0-riscv64-unknown-linux-gnu/lib/python3.12/site-packages/bcrypt/_bcrypt.cpython-312-riscv64-linux-gnu.so
/nix/store/xfc8xssvq2vwcn6126ml3m6fkdd1gzw0-python3.12-bcrypt-4.2.0-riscv64-unknown-linux-gnu/lib/python3.12/site-packages/bcrypt/_bcrypt.cpython-312-riscv64-linux-gnu.so: ELF 64-bit LSB shared object, x86-64, version 1 (SYSV), dynamically linked, not stripped
| 07:34:49 |
| DavHau | The binary is a result from a rust build. The derivation in nixpks looks fine. Other python packages containing rust cross compile correctly | 07:36:48 |
 Alex | The Python package's install script might not be cross-friendly; you should look at the package's source.
If it's doing the build itself, it probably found the buildPackages compiler instead of the cross compiler. | 07:40:29 |
| DavHau | OK, for rust the compiler is target platform specific like with gcc? | 11:10:44 |
| Alex | Oh I just noticed it has the riscv64 target triple in the file name, very weird.
No, rustc is multi-targeting because it's based on LLVM. It's probably being given the wrong build flags.
Have you looked through the build logs for anything suspicious? | 11:45:27 |
| DavHau | Could not spot anything. I guess I'll inspect the source once I find the time. | 16:40:25 |
| 26 Dec 2024 |
|  rtunreal changed their display name from rtunreal to rtunreal | SIP:3987. | 14:26:42 |
| 27 Dec 2024 |
|  root left the room. | 06:51:45 |
|  raitobezarius changed their display name from raitobezarius to raitobezarius (DECT: 3538 / EPVPN 2681). | 07:31:40 |
|  pinpox changed their display name from pinpox to pinpox [DECT: 7170]. | 15:23:12 |
| 30 Dec 2024 |
| raitobezarius changed their display name from raitobezarius (DECT: 3538 / EPVPN 2681) to raitobezarius. | 16:27:14 |
| 1 Jan 2025 |
|  NixOS Moderation Botchanged room power levels. | 14:26:06 |
| pinpox changed their display name from pinpox [DECT: 7170] to pinpox. | 18:20:16 |
|  jopejoe1 (4094@epvpn) changed their display name from jopejoe1 [4094] to jopejoe1. | 23:29:29 |
|  scm left the room. | 23:53:47 |
| 4 Jan 2025 |
 Tristan Ross | Why does nixos-hardware use the upstream u-boot instead of the Starfive fork for the VisionFive 2? I'm looking to boot from an NVME and it looks like using the fork would be better. | 15:59:03 |
| Alex | In reply to @rosscomputerguy:matrix.org
Why does nixos-hardware use the upstream u-boot instead of the Starfive fork for the VisionFive 2? I'm looking to boot from an NVME and it looks like using the fork would be better. Upstream is usually a better choice when the hardware support you need has been merged. For extra hardware, applying patches onto upstream is a decent option because it helps test the patches before they get merged. | 18:09:33 |
| Tristan Ross | Yeah, I made this PR https://github.com/NixOS/nixos-hardware/pull/1296 to make overriding those things easier | 18:10:18 |
| Tristan Ross | After running into a few things which don't like RISC-V or cross compiling (lsd, nix output monitor, Prism Launcher, and lua-language-server), I seem to be making good progress. | 19:04:27 |
| Alex | In reply to @rosscomputerguy:matrix.org
After running into a few things which don't like RISC-V or cross compiling (lsd, nix output monitor, Prism Launcher, and lua-language-server), I seem to be making good progress. nix-output-monitor can be compiled if you can provide Nixpkgs a boot GHC | 21:00:11 |
| Tristan Ross | lol, I'm too lazy to go through that effort | 21:00:33 |
| Tristan Ross | I'm just trying to get my config to build for the first time via cross from aarch64 to riscv64 | 21:00:54 |
| Tristan Ross | Also, the VisionFive 2 isn't that powerful so not like I'm going to be doing much on it | 21:01:25 |
| 5 Jan 2025 |
| Tristan Ross | Successful build of my config for the VF2, time to try and get it building with COSMIC. | 02:33:37 |
| Tristan Ross | 330 builds | 02:35:29 |
| Tristan Ross | I couldn't get COSMIC but I did get my system to build and boot | 05:52:25 |
| Tristan Ross | jegan% lsblk
NAME MAJ:MIN RM SIZE RO TYPE MOUNTPOINTS
mtdblock0 31:0 0 960K 0 disk
mtdblock1 31:1 0 64K 0 disk
mtdblock2 31:2 0 15M 0 disk
mmcblk1 179:0 0 238.3G 0 disk
Ãmmcblk1p1 179:1 0 2M 0 part
├─mmcblk1p2 179:2 0 4M 0 part
└─mmcblk1p3 179:3 0 238.3G 0 part /nix/store
/
nvme0n1 259:0 0 931.5G 0 disk
Looks like the NVME is present
| 05:52:39 |
| Alex | In reply to @rosscomputerguy:matrix.org
jegan% lsblk
NAME MAJ:MIN RM SIZE RO TYPE MOUNTPOINTS
mtdblock0 31:0 0 960K 0 disk
mtdblock1 31:1 0 64K 0 disk
mtdblock2 31:2 0 15M 0 disk
mmcblk1 179:0 0 238.3G 0 disk
Ãmmcblk1p1 179:1 0 2M 0 part
├─mmcblk1p2 179:2 0 4M 0 part
└─mmcblk1p3 179:3 0 238.3G 0 part /nix/store
/
nvme0n1 259:0 0 931.5G 0 disk
Looks like the NVME is present
NB: this doesn't mean it's bootable. | 07:53:46 |
| Tristan Ross | In reply to @alex:tunstall.xyz
NB: this doesn't mean it's bootable. I know, I'll need to pull in the Starfive OpenSBI and u-boot according to the forums. | 14:42:25 |
| Tristan Ross | It's at least progress, when I tried this in the past it wasn't even showing up in lsblk. | 14:43:00 |
