| 3 Feb 2026 |
 Randy Eckenrode | I take it the test doesn’t use the existing test runner? I ran into this with Swift’s LLVM too. | 15:04:13 |
 niklaskorz | The test itself isn't the problem, but it tries to execute a binary with an empty environment on purpose | 15:18:39 |
| niklaskorz | so that binary will need to receive DYLD_LIBRARY_PATH somehow | 15:18:52 |
| niklaskorz | we could patch https://github.com/llvm/llvm-project/blob/d6211b112fa75fb5aa0bcd3885d3c9c36f4deece/llvm/unittests/Support/ProgramTest.cpp#L695-L696 to pass just DYLD_LIBRARY_PATH instead of ArrayRef<StringRef>{} | 15:22:12 |
| Randy Eckenrode | How does it work on Linux? Does Linux not need LD_LIBRARY_PATH, or does it set the rpath on the binary? | 16:29:20 |
| niklaskorz | In reply to @reckenrode:matrix.org
How does it work on Linux? Does Linux not need LD_LIBRARY_PATH, or does it set the rpath on the binary? Good question, I'll check tomorrow | 20:43:15 |
| 4 Feb 2026 |
|  tiferrei changed their profile picture. | 03:03:20 |
| tiferrei left the room. | 14:28:35 |
| niklaskorz | on hold because I reproducibly end up with a segmentation fault when trying to build LLVM 22 on Linux 🤔 | 16:09:49 |
| niklaskorz | but it passes on hydra, so I'll try another machine | 16:10:27 |
| niklaskorz | still inclined to just skip this test on darwin as it was only added to test a windows quirk | 16:13:59 |
| 5 Feb 2026 |
|  jarver joined the room. | 06:02:51 |
|  Achmad Mahardi joined the room. | 07:46:53 |
| 6 Feb 2026 |
|  samemrecebi joined the room. | 00:08:29 |
 vcunat | Hi. We have weird regressions on aarch64-darwin again. In log I see
XprotectService: [com.apple.xprotect:xprotect] File /nix/store/m5q1xl33fmnmxrfmipq8aqzlr8f0b4ln-texlive-bin-2025-kpathsea/bin/kpsewhich
failed on loadCmd /nix/store/8829nnvx1kfrv63qc0yv838ppwhmny35-texlive-bin-2025/lib/libkpathsea.6.dylib (loadCmd resolved to: /nix/store/8829nnvx1kfrv63qc0yv838ppwhmny35-texlive-bin-2025/lib/libkpathsea.6.dylib, bundleURL:
(null))
| 11:05:04 |
| vcunat | Is that - what was it - non-deterministic failures to correctly sign the exe/lib? | 11:06:31 |
| vcunat | * Is that - what was it - non-deterministic failures to correctly sign the exe/lib again? | 11:06:35 |
| vcunat | (for reference, this is on the current staging-next-25.11 branch, and x86_64-darwin seems fine on the same build) | 11:07:49 |
| vcunat | Apparently it can be worked around by changing the hash to force a rebuild (of the successful texlive-bin-2025 build), so let me do that for now. | 11:24:20 |
| vcunat | * Apparently it can be worked around by changing the hash to force a rebuild (of the successful texlive-bin-2025 build), so let me do that for now in 1e6c61a5a473. | 11:27:06 |
 denommus | is cross-compilation on macos broken? | 17:19:57 |
 K900 | To what target? | 17:23:19 |
| denommus | I want to compile something to x86_64-linux | 17:23:23 |
| denommus | I'm on a M3 | 17:23:30 |
| denommus | glibc fails to build because of a broken patch | 17:23:50 |
| K900 | Yeah very unlikely | 17:23:56 |
| K900 | You probably want to set up a Linux VM | 17:24:02 |
| K900 | And build on that | 17:24:03 |
| Randy Eckenrode | Cross to Linux should work. | 17:26:45 |
| denommus | I think even if I set up a Linux VM I'll need to crosscompile, because the VM would be aarch64 as well, since I'm on a M3 | 17:27:41 |
