| 6 Jan 2025 |
 Pratham Patel (you can mention me) | You can cross compile ghc for rv64 but cannot bootstrap it natively (yet) | 03:29:24 |
 Tristan Ross | Yeah, I was looking at the issue | 03:34:49 |
| Tristan Ross | The problem is there's code to disable shellcheck on incompatible problems | 03:35:04 |
| Tristan Ross | And the problem is that check does not work | 03:35:12 |
| Tristan Ross | nixos-unstable should have it, I pointed the config to that and yet it still errors | 03:38:33 |
| Tristan Ross | Ok, master branch works | 03:46:50 |
| Tristan Ross | Wait, is the nixpkgs issue tracker wrong lol? It says the PR made it through fully | 03:48:47 |
| Tristan Ross | Yay, we've gotten passed eval. Now time for like 2 weeks of compiling. | 03:55:44 |
|  spd joined the room. | 15:40:46 |
| 7 Jan 2025 |
|  xin joined the room. | 07:42:18 |
|  Sergei Zimmerman (xokdvium) joined the room. | 11:41:55 |
 Alex | In reply to @rosscomputerguy:matrix.org
And the problem is that check does not work I "broke" the check when I added support for cross-compiling GHC. All you need to do is override the default GHC to disable building it for the RISC-V platform such that the check functions as originally intended. | 22:22:06 |
| Tristan Ross | In reply to @alex:tunstall.xyz
I "broke" the check when I added support for cross-compiling GHC. All you need to do is override the default GHC to disable building it for the RISC-V platform such that the check functions as originally intended. Oh... fun | 22:25:02 |
| Tristan Ross | Well it's been building for 50 hours at this point so it's doing fine | 22:25:23 |
| Tristan Ross | I estimate another 40 hours to go | 22:26:55 |
| 9 Jan 2025 |
| Tristan Ross | error: builder for '/nix/store/ra18qbxfgr6a2ydln4ja4i56qqql9jai-bpftools-6.12.drv' failed with exit code 1;
last 25 log lines:
> Auto-detecting system features:
> ... libbfd: [ on ]
>
> DESCEND bpftool
>
> Auto-detecting system features:
> ... clang-bpf-co-re: [ OFF ]
> ... llvm: [ OFF ]
> ... libcap: [ OFF ]
> ... libbfd: [ on ]切断 | ttyUSB0 >
> MKDIR /build/linux-6.12/tools/bpf/bpftool/libbpf/
> GEN /build/linux-6.12/tools/bpf/bpftool/libbpf/bpf_helper_defs.h
> CC /build/linux-6.12/tools/bpf/bpftool/libbpf/staticobjs/libbpf.o
> In file included from /build/linux-6.12/tools/include/asm/barrier.h:12,
> from /build/linux-6.12/tools/include/linux/ring_buffer.h:4,
> from libbpf.c:37:
> /build/linux-6.12/tools/include/asm/../../arch/riscv/include/asm/barrier.h:13:10: fatal error: asm/fence.h: No such file or directory
> 13 | #include <asm/fence.h>
> | ^~~~~~~~~~~~~
> compilation terminated.
> make[3]: *** [/build/linux-6.12/tools/build/Makefile.build:106: /build/linux-6.12/tools/bpf/bpftool/libbpf/staticobjs/libbpf.o] Error 1
> make[2]: *** [Makefile:164: /build/linux-6.12/tools/bpf/bpftool/libbpf/staticobjs/libbpf-in.o] Error 2
> make[1]: *** [Makefile:46: /build/linux-6.12/tools/bpf/bpftool/libbpf/libbpf.a] Error 2
> make: *** [Makefile:113: bpftool] Error 2
For full logs, run 'nix log /nix/store/ra18qbxfgr6a2ydln4ja4i56qqql9jai-bpftools-6.12.drv'.
Fun
| 17:57:13 |
| Tristan Ross | https://github.com/NixOS/nixpkgs/pull/371240 oh | 17:57:47 |
| Alex |
切断 | ttyUSB0
This isn't actually in the build logs, is it? ("Disconnected" in ja_JP locale?) | 18:02:23 |
| Tristan Ross | That's Chinese | 18:05:36 |
| Tristan Ross | And this is over serial so I think there's some clobbering of data | 18:05:46 |
| Tristan Ross | Idk how Chinese text got in aside from some earlier logs from lua-language-server failing to build and it prints its unit tests in Chinese | 18:06:59 |
| Tristan Ross | I pulled in that bpftools fix into my Flake and looks like it's working? I'm not sure until I have significant progress in this rerun of the builds. | 18:08:13 |
| Tristan Ross | 592 builds to go | 18:08:20 |
| Alex | In reply to @rosscomputerguy:matrix.org
That's Chinese Impossible to tell them apart because of the way Unicode was designed. | 19:04:42 |
| Tristan Ross | Rip, openexr failed tests due to 1 test timing out | 19:45:13 |
| Alex | In reply to @rosscomputerguy:matrix.org
Rip, openexr failed tests due to 1 test timing out This unfortunately happens for some packages. Commonly available RISC-V cores can be far slower than the heavily optimised and mature x86 platforms that often set the timings. You could report the issue upstream, but for now disabling checks is an easy solution. It could even save hours of building if you do it pervasively. | 21:20:48 |
| Tristan Ross | In reply to @alex:tunstall.xyz
This unfortunately happens for some packages. Commonly available RISC-V cores can be far slower than the heavily optimised and mature x86 platforms that often set the timings. You could report the issue upstream, but for now disabling checks is an easy solution. It could even save hours of building if you do it pervasively. Yeah... I'm kinda aware of that. Hopefully someday we get more performant systems. | 21:22:08 |
| Alex | Emulated builds (e.g. via binfmt) also struggle with the problem. Access to clocks and timings interferes with reproducibility, but they are very difficult system calls to take away. | 21:24:54 |
| Tristan Ross | Unfortunately, I cannot do emulated builds because those don't work with a page size larger than 4k. | 23:13:05 |
| Tristan Ross | I've gotten things like bus errors with just running ls | 23:13:20 |
