| 4 Jul 2026 |
emily | ok, fixed pikepdf | 13:59:17 |
emily |
pendulum is broken on Linux and also downstream of pikepdf
django was flaky and is building now
- Swift stuff I need to look at
ldacBT looks like weird dumb stuff but I'll look at it briefly
qtdeclarative is… downstream of pikepdf somehow?
poppler-glib is the same crap as pikepdf so should be an easy fix
opencv is the ld64 crap
syrupy is… hexa wtf is this https://hydra.nixos.org/build/334176737/nixlog/1
| 14:03:42 |
hexa | a lot of whitespace | 14:03:57 |
emily | I am not seeing a failure | 14:04:07 |
emily | I'll fix everything except Swift, then spend a little more time banging my head against Swift, and then I think I'm done for this cycle… everything below those blocks <100 builds so I don't care | 14:04:37 |
hexa | it does build for me | 14:04:45 |
emily | well done except for the x86_64-darwin drop stuff 🙃 | 14:04:50 |
emily | ok, it was blocked on matplotlib beforehand. weird flaky nonsense? | 14:05:06 |
emily | I'll hit restart | 14:05:31 |
emily | so ldacbt was almost certainly e1e3ccda5d37e1796efbfbe2ca1b7e2887e17ab1 I guess | 14:13:54 |
emily | K900: wasn't there some screwy licensing thing with this encoding stuff | 14:14:36 |
K900 | What encoding stuff | 14:14:48 |
emily | the LDAC stuff | 14:15:10 |
emily | some weird source dump thing we made opt-in | 14:15:16 |
emily | # Upstream ships AOSP build files and a gcc/ makefile that only knows
# about the in-tree layout. Compile and link directly; the entire
# library is two umbrella translation units.
buildPhase = ''
runHook preBuild
soname=libldacBT.so.${lib.versions.major finalAttrs.version}
sofile=libldacBT.so.${finalAttrs.version}
$CC -shared -Wl,-soname,$soname src/ldaclib.c src/ldacBT.c abr/src/ldacBT_abr.c -lm -o $sofile
runHook postBuild
'';
| 14:15:38 |
K900 | It's not licensing | 14:15:38 |
Alyssa Ross | yeah can do | 14:15:42 |
K900 | The build system is screwy | 14:15:43 |
emily | so this PR obviously just broke the build completely on Darwin which doesn't use .so or -soname or any of that | 14:15:50 |
K900 | But it's all apache2 in Sony or someone's AOSP drop | 14:15:53 |
K900 | Why the fuck does anything use ldacbt on Darwin | 14:16:05 |
emily | moving from CMake to a raw compiler invocation that also uses env.NIX_CFLAGS_COMPILE unnecessarily on top of that is not very funny | 14:16:22 |
emily | GStreamer etc. | 14:16:27 |
K900 | Oh god | 14:16:33 |
K900 | Why does gstreamer | 14:16:38 |
K900 | Ugh | 14:16:39 |
emily | I'm a bit annoyed at this PR being merged | 14:17:11 |
emily | like we went from a normal CMake build to manually calling the compiler and writing out a bunch of .pc files ourselves | 14:17:30 |
K900 | Sorry I assumed it was only used by pipewire | 14:17:48 |
K900 | Apparently not the case | 14:17:53 |