| 7 Mar 2026 |
dramforever | but it is passing different -march to different objects | 11:45:39 |
dramforever | so something is going on | 11:45:43 |
K900 | That postPatch line is also hella sus | 11:46:00 |
K900 | In the nixpkgs PR change | 11:46:06 |
K900 | In reply to @yuka:yuka.dev why would my different build hosts fail with different errors then? Could just be ordering | 11:46:10 |
dramforever |
target specific option mismatch
maybe that was it, maybe openblas shouldn't be doing it
| 11:46:19 |
K900 | No, I think that's fine | 11:46:30 |
K900 | Conceptually | 11:46:32 |
K900 | Like that's not a "you're trying to link objects with different march" error | 11:47:11 |
dramforever | on a second thought yeah | 11:47:22 |
K900 | That's a "you're trying to call stuff your march doesn't have" error | 11:47:24 |
Yureka (she/her) | I'm trying one very easy thing now, which is building latest OpenBLAS develop | 11:48:30 |
Yureka (she/her) | nope, also broken on M1 Pro | 11:48:56 |
K900 | I wonder if there is a "just stop trying to fucking guess" flag somewhere | 11:51:56 |
Yureka (she/her) | Ah no, actually the log you saw with the neoverse march was from the M1 Pro | 11:53:09 |
Yureka (she/her) | so that was not guessed | 11:53:24 |
Yureka (she/her) | so maybe that's fine | 11:53:29 |
dramforever | #450106 was supposedly test built on aarch64-linux | 11:54:06 |
dramforever | so it's at least fragile somehow? | 11:54:18 |
Yureka (she/her) | https://github.com/openMathLib/openBLAS/commit/05fc88180c42eae7588fac4bd59c3b86ffefbc23 | 12:10:18 |
Yureka (she/her) | reverting this fixes the build on both Ampere and M1 Pro | 12:10:33 |
Yureka (she/her) | https://github.com/NixOS/nixpkgs/pull/497538 | 12:15:41 |
K900 | Can't you just makeflags BUILD_BFLOAT16=0? | 12:21:56 |
dramforever | oh come on
postPatch = ''
# cc1: error: invalid feature modifier 'sve2' in '-march=armv8.5-a+sve+sve2+bf16'
substituteInPlace Makefile.arm64 --replace-fail "+sve2+bf16" ""
'';
| 12:23:04 |
dramforever | come on | 12:23:07 |
dramforever | turns out we played ourselves | 12:23:53 |
dramforever | commented on the pr | 12:28:34 |
Yureka (she/her) | wait what? | 12:36:24 |
Yureka (she/her) | wouldn't replace-fail fail if it didn't replace anything? | 12:36:34 |
Yureka (she/her) | or does it only replace the first occurance | 12:38:36 |