Rerender

[h-vetinari]

Due to changes in rattler, the recipe is currently not rerenderable anymore. Also the migrator file was named wrongly in #12. Fix both, so that the bot can handle this feedstock again.

[conda-forge-admin]

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe/recipe.yaml) and found it was in an excellent condition.

I do have some suggestions for making it better though...

For recipe/recipe.yaml:

• ℹ️ The use of xz is deprecated. If your package links to liblzma, use the new liblzma-devel package. For XZ's command line utilities (like xz and unxz), use the xz-toolspackage. For the xz-enabled GNU tools (like grep), use xz-gpl-tools. Note: Command line utilities should go in requirements/build.

_{This message was generated by GitHub Actions workflow run https://github.com/conda-forge/conda-forge-webservices/actions/runs/14051844122. Examine the logs at this URL for more detail.}

[h-vetinari]

OK, this is now segfaulting consistently for CUDA 12.6 in what appears to be the kaldi integration tests

 │ test/torchaudio_unittest/functional/kaldi_compatibility_cpu_test.py::TestKaldiFloat32::test_sliding_window_cmn PASSED [ 65%]
 │ test/torchaudio_unittest/functional/kaldi_compatibility_cpu_test.py::TestKaldiFloat64::test_sliding_window_cmn PASSED [ 65%]
 │ test/torchaudio_unittest/functional/kaldi_compatibility_cuda_test.py::TestKaldiFloat32::test_sliding_window_cmn SKIPPED [ 65%]
 │ test/torchaudio_unittest/functional/kaldi_compatibility_cuda_test.py::TestKaldiFloat64::test_sliding_window_cmn SKIPPED [ 65%]
 │ Fatal Python error: Segmentation fault
 │ Current thread 0x00007f291a033400 (most recent call first):

I suspect that we should rebuild kaldi for 12.6 but this has failed: conda-forge/kaldi-feedstock#53

[h-vetinari]

Well, the kaldi connection was a red herring. Turns out there's only 4 tests in that blocked (I had assumed there's more; here from a passing run on CUDA 11.8)

 │ test/torchaudio_unittest/functional/kaldi_compatibility_cpu_test.py::TestKaldiFloat32::test_sliding_window_cmn PASSED [ 65%]
 │ test/torchaudio_unittest/functional/kaldi_compatibility_cpu_test.py::TestKaldiFloat64::test_sliding_window_cmn PASSED [ 65%]
 │ test/torchaudio_unittest/functional/kaldi_compatibility_cuda_test.py::TestKaldiFloat32::test_sliding_window_cmn SKIPPED [ 65%]
 │ test/torchaudio_unittest/functional/kaldi_compatibility_cuda_test.py::TestKaldiFloat64::test_sliding_window_cmn SKIPPED [ 65%]
 │ test/torchaudio_unittest/functional/librosa_compatibility_cpu_test.py::TestFunctionalCPU::test_amplitude_to_DB PASSED [ 65%]
 │ test/torchaudio_unittest/functional/librosa_compatibility_cpu_test.py::TestFunctionalCPU::test_amplitude_to_DB_power PASSED [ 65%]

which means that the segfaulting test is actually librosa_compatibility_cpu_test.py::TestFunctionalCPU::test_amplitude_to_DB. However, librosa has been on 0.10.2.post1 both in the failing run here, as well as the last passing run on main. I also don't see what pure-python package like the (horribly-named) librosa could have to do with a segfault...

[h-vetinari]

Ah... Just realized that it's (most likely) also due to conda-forge/llvmlite-feedstock#99 😑

 │ Fatal Python error: Segmentation fault
 │ Current thread 0x00007f4b018db400 (most recent call first):
 │   File "$PREFIX/lib/python3.11/site-packages/llvmlite/binding/ffi.py", line 197 in __call__
 │   File "$PREFIX/lib/python3.11/site-packages/llvmlite/binding/executionengine.py", line 99 in finalize_object
 │   File "$PREFIX/lib/python3.11/site-packages/numba/core/codegen.py", line 1068 in wrapper
 │   [...]
 │   File "$PREFIX/lib/python3.11/site-packages/librosa/core/audio.py", line 1158 in <module>

[h-vetinari]

llvmlite is finally fixed, now needs conda-forge/kaldi-feedstock#56