24 Jul 2024 |
rhelmot | the locked build seems to have been garbage collected, or perhaps I'm using nix wrong :) I'll take this as an opportunity to re-lock it with the up-to-date stuff | 17:24:16 |
John Ericson | rhelmot: ok great! | 17:25:42 |
| @ftchrist:matrix.org left the room. | 19:35:44 |
25 Jul 2024 |
dramforever | In reply to @trofi:matrix.org worth fixing qemu instead if feasible? there's a lot, like most of the time it's unimplemented features and sometimes it's bugs. archlinux riscv build most of their stuff on qemu-user and they have a list of failures here: https://github.com/felixonmars/archriscv-packages/blob/master/qemu-user-blacklist.txt | 00:31:01 |
dramforever | it's often test suites though | 00:31:29 |
dramforever | fixing qemu would work for some cases like i did manage to fix the coreutils problem but i doubt it would be generally applicable | 00:39:53 |
dramforever | * fixing qemu would work for some cases like i did manage to fix the coreutils problem but i doubt it would be generally applicable/feasible | 00:39:59 |
28 Jul 2024 |
| matthewcroughan changed their display name from matthewcroughan to matthewcroughan - going to nix.camp. | 16:15:40 |
artemist | Is there a reason we build compiler-rt twice on FreeBSD? Upstream clang hardcodes `-lgcc --as-needed -lgcc_s --no-as-needed` link arguments so the `libclang_rt.builtins` library should never be used | 17:52:02 |
artemist | We provide both libgcc.a and libgcc_s.so in freebsd.libc and they're better on some architectures because they're built with FreeBSD headers, e.g. they're required unless we patch out CPU feature detection on aarch64-freebsd | 18:03:30 |
30 Jul 2024 |
John Ericson | @artemist:mildlyfunctional.gay: oops I thought I added a thing so we only built it once? | 17:30:01 |
John Ericson | Before I merged the PR? | 17:30:31 |
John Ericson | Also
```
/build/compiler-rt-src-18.1.8/compiler-rt/lib/builtins/cpu_model/x86.c:900:3: error: call to undeclared function 'static_assert'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration]
900 | static_assert(sizeof(Features) / sizeof(Features[0]) == 4, "");
| ^
/build/compiler-rt-src-18.1.8/compiler-rt/lib/builtins/cpu_model/x86.c:939:3: error: call to undeclared function '__assert'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration]
939 | assert(__cpu_model.__cpu_vendor < VENDOR_MAX);
| ^
/build/compiler-rt-src-18.1.8/compiler-rt/lib/builtins/cpu_model/x86.c:24:32: note: expanded from macro 'assert'
24 | #define assert(e) ((e)?(void)0:__assert(__FUNCTION__,__FILE__,__LINE__,#e))
|
``` | 17:30:43 |
John Ericson | I noticed this failure building freebsd static, I think it's new because got almost the same failure with OpenBSD and that was working a few weeks ago | 17:32:01 |
John Ericson | (also with LLVM 18) | 17:32:09 |
artemist | We might just want to add FreeBSD headers to compiler-rt, then remove it from freebsd.libc | 18:36:56 |
artemist | We'd need to add another link to libcompiler_rt.a , but that's not a huge problem | 18:37:16 |
John Ericson | artemist: well with the current tricks we might already be using the LLD cyclic linking? | 18:48:41 |
John Ericson | I will try to see what changed in recently | 18:48:48 |
John Ericson | It is not good to rely on undefined symbols, but I am curious | 18:49:05 |
artemist | static_assert is defined in FreeBSD headers to _Static_assert , which is a clang keyword, so it shouldn't mean anything for linking | 18:53:31 |
artemist | I've mostly been messing with compiler-rt becuase libc on aarch64-freebsd needs compiler-rt, but compiler-rt doesn't require libc | 18:54:16 |
John Ericson | artemist: will I see your compiler-rt work if I go look at open PRs? | 18:58:04 |
John Ericson | ...I guess that ' | 18:58:13 |
John Ericson | * ...I guess that's something I can swner myself ;) | 18:58:19 |
John Ericson | * ...I guess that's something I can answer myself :) | 18:58:26 |
artemist | No, it's part of aarch64 stuff that needs cleanup | 18:58:31 |
artemist | I can try to split it into its own PR today though | 18:59:11 |
John Ericson | OK | 18:59:57 |
John Ericson | I just saw your 1 open PR and merged it, btw | 19:00:04 |