| 15 Jan 2022 |
 Zhaofeng Li | In reply to @zhaofeng:zhaofeng.li
Vanilla Firefox is buildable after a simple patch to authenticator-rs to add RISC-V support. I wasn't able to build node.js natively (stuck on node_mksnapshot) which is used in the build process, so I cheated a little bit by using the X86 version of node.js. (Actually built through binfmt on x86, since I did the nodejs trick 😛) | 18:30:24 |
 hexa | i'd rather have riscv support in common.nix than darwin fwiw 😀 | 18:30:59 |
 [0x4A6F] | About binfmt, how did you set this up for riscv64-linux? | 18:41:01 |
| Zhaofeng Li | Just boot.binfmt.emulatedSystems = [ "riscv64-linux" ]; works fine | 18:41:33 |
| [0x4A6F] | Doesn't work for one of my system. still investigating. aarch64-linux works though. | 18:43:07 |
| Zhaofeng Li | Do you see the binfmt_misc registrations in /proc/sys/fs/binfmt_misc at all? | 18:44:13 |
| [0x4A6F] | Just a problem in my deployment..Works now. | 18:54:30 |
 sterni | Zhaofeng Li: so a downstream native stdenv? | 19:10:00 |
| Zhaofeng Li | In reply to @sternenseemann:systemli.org
Zhaofeng Li: so a downstream native stdenv? Not anymore, since riscv64-linux stdenv got merged in https://github.com/NixOS/nixpkgs/pull/147942 | 19:12:01 |
| sterni | uhh missed that nice :) | 19:44:01 |
| 16 Jan 2022 |
| Zhaofeng Li | In case you want to try out Firefox and SuperTuxKart on RISC-V NixOS yourself, there is now a new graphical QEMU image:
https://github.com/zhaofengli/nixos-riscv64/releases/tag/2022011500 | 04:58:11 |
| Zhaofeng Li | With Virtio-GPU + VirGL (use the included launch.sh), SuperTuxKart is pretty playable with some texture glitches | 04:59:44 |
 ius | Zhaofeng Li: re binfmt builds, have you run into any qemu issues? boehm-gcs testsuite seems to ~reliably cause qemu-user to disintegrate | 22:02:31 |
| ius | ERROR:../plugins/core.c:222:qemu_plugin_vcpu_init_hook: assertion failed: (success)
Bail out! ERROR:../plugins/core.c:222:qemu_plugin_vcpu_init_hook: assertion failed: (success)
Child process failed, status= 0x7f00
Test failed
FAIL gctest (exit status: 134)
(on nixos-unstable, riscv64)
| 22:03:07 |
| Zhaofeng Li | In reply to @ius:nltrix.net
ERROR:../plugins/core.c:222:qemu_plugin_vcpu_init_hook: assertion failed: (success)
Bail out! ERROR:../plugins/core.c:222:qemu_plugin_vcpu_init_hook: assertion failed: (success)
Child process failed, status= 0x7f00
Test failed
FAIL gctest (exit status: 134)
(on nixos-unstable, riscv64)
Hmm, binfmt is pretty janky and it could be one of the packages that don't work. Let me try from my end | 22:04:15 |
| ius | looks like some kind of threading issue, and I'm not the first to notice: https://lists.gnu.org/archive/html/qemu-devel/2020-04/msg03004.html | 22:07:41 |
| Zhaofeng Li | Hmm, it builds fine for me (qemu 6.1.0): https://gist.github.com/zhaofengli/95eb2948deda98b6d8f2854bdf24c24f | 22:08:27 |
| Zhaofeng Li | (this is under binfmt) | 22:08:38 |
| ius | Same qemu ver for me, failed 3 times in a row. | 22:24:11 |
| ius | Sometimes on a different test, so I guess there's some bad luck involved ;) | 22:24:33 |
| ius | Well, patching the assert out works for me 🤷♂️ | 22:36:24 |
| 18 Jan 2022 |
| ius | Zhaofeng Li: I've built some more and see similar wonkyness in aws-common-c and llvm tests - are you running those tests & are they okay for you using binfmt? | 19:26:33 |
| ius | I'm confused how all these tests suggest it's extremely broken, yet the toolchain seems fine | 19:31:33 |
| ius | 43% tests passed, 222 tests failed out of 389 | 19:34:31 |
| Zhaofeng Li | For aws-c-common, see https://github.com/zhaofengli/nixpkgs/commit/4f80543ea818b0e833756619d8484d245ec7477a and https://github.com/awslabs/aws-c-common/pull/873 | 19:36:09 |
| Zhaofeng Li | And for LLVM, the tests need to be disabled for the time being. Some of the issues are being tracked upstream. | 19:36:58 |
| ius | Thanks, again | 23:39:30 |
| 20 Jan 2022 |
|  andi- left the room. | 00:09:30 |
| Zhaofeng Li | In reply to @ius:nltrix.net
Zhaofeng Li: re binfmt builds, have you run into any qemu issues? boehm-gcs testsuite seems to ~reliably cause qemu-user to disintegrate Ok, I just rebased my branch against master and can confirm that I see the same thing | 02:12:44 |
| ius | Let me know if you figure out why, otherwise I might look at it (maybe in a bit) | 17:38:34 |
