Build failure: audio/torchaudio/csrc/decoder/src/decoder/Utils.h:23:40: error: ‘numeric_limits’ is not a member of ‘std’

[kwand]

🐛 Describe the bug

I encountered the following build error after cloning the master branch and attempting to build/install it using CC=gcc-11 CXX=g++-11 python setup.py install --user

[3/44] /usr/bin/g++-11 -DBUILD_CTC_DECODER -DHAVE_BZLIB -DHAVE_XZLIB -DHAVE_ZLIB -DINCLUDE_KALDI -DINCLUDE_SOX -DKENLM_MAX_ORDER=6 -DUSE_C10D_GLOO -DUSE_C10D_MPI -DUSE_C10D_NCCL -DUSE_CUDA -DUSE_DISTRIBUTED -DUSE_RPC -DUSE_TENSORPIPE -Dlibtorchaudio_decoder_EXPORTS -I/home/dkwan/Github/Builds/audio -I/home/dkwan/Github/Builds/audio/third_party/kenlm/submodule -I/home/dkwan/Github/Builds/audio/third_party/zlib/../install/include -I/home/dkwan/Github/Builds/audio/third_party/bzip2/../install/include -I/home/dkwan/Github/Builds/audio/third_party/lzma/../install/include -isystem /home/dkwan/.local/lib/python3.10/site-packages/torch/include -isystem /home/dkwan/.local/lib/python3.10/site-packages/torch/include/torch/csrc/api/include -isystem /opt/cuda/include -Wall -D_GLIBCXX_USE_CXX11_ABI=1 -O3 -DNDEBUG -fPIC -D_GLIBCXX_USE_CXX11_ABI=1 -std=gnu++14 -MD -MT torchaudio/csrc/CMakeFiles/libtorchaudio_decoder.dir/decoder/src/decoder/LexiconFreeDecoder.cpp.o -MF torchaudio/csrc/CMakeFiles/libtorchaudio_decoder.dir/decoder/src/decoder/LexiconFreeDecoder.cpp.o.d -o torchaudio/csrc/CMakeFiles/libtorchaudio_decoder.dir/decoder/src/decoder/LexiconFreeDecoder.cpp.o -c /home/dkwan/Github/Builds/audio/torchaudio/csrc/decoder/src/decoder/LexiconFreeDecoder.cpp
FAILED: torchaudio/csrc/CMakeFiles/libtorchaudio_decoder.dir/decoder/src/decoder/LexiconFreeDecoder.cpp.o 
/usr/bin/g++-11 -DBUILD_CTC_DECODER -DHAVE_BZLIB -DHAVE_XZLIB -DHAVE_ZLIB -DINCLUDE_KALDI -DINCLUDE_SOX -DKENLM_MAX_ORDER=6 -DUSE_C10D_GLOO -DUSE_C10D_MPI -DUSE_C10D_NCCL -DUSE_CUDA -DUSE_DISTRIBUTED -DUSE_RPC -DUSE_TENSORPIPE -Dlibtorchaudio_decoder_EXPORTS -I/home/dkwan/Github/Builds/audio -I/home/dkwan/Github/Builds/audio/third_party/kenlm/submodule -I/home/dkwan/Github/Builds/audio/third_party/zlib/../install/include -I/home/dkwan/Github/Builds/audio/third_party/bzip2/../install/include -I/home/dkwan/Github/Builds/audio/third_party/lzma/../install/include -isystem /home/dkwan/.local/lib/python3.10/site-packages/torch/include -isystem /home/dkwan/.local/lib/python3.10/site-packages/torch/include/torch/csrc/api/include -isystem /opt/cuda/include -Wall -D_GLIBCXX_USE_CXX11_ABI=1 -O3 -DNDEBUG -fPIC -D_GLIBCXX_USE_CXX11_ABI=1 -std=gnu++14 -MD -MT torchaudio/csrc/CMakeFiles/libtorchaudio_decoder.dir/decoder/src/decoder/LexiconFreeDecoder.cpp.o -MF torchaudio/csrc/CMakeFiles/libtorchaudio_decoder.dir/decoder/src/decoder/LexiconFreeDecoder.cpp.o.d -o torchaudio/csrc/CMakeFiles/libtorchaudio_decoder.dir/decoder/src/decoder/LexiconFreeDecoder.cpp.o -c /home/dkwan/Github/Builds/audio/torchaudio/csrc/decoder/src/decoder/LexiconFreeDecoder.cpp
In file included from /home/dkwan/Github/Builds/audio/torchaudio/csrc/decoder/src/decoder/Decoder.h:10,
                 from /home/dkwan/Github/Builds/audio/torchaudio/csrc/decoder/src/decoder/LexiconFreeDecoder.h:12,
                 from /home/dkwan/Github/Builds/audio/torchaudio/csrc/decoder/src/decoder/LexiconFreeDecoder.cpp:14:
/home/dkwan/Github/Builds/audio/torchaudio/csrc/decoder/src/decoder/Utils.h:23:40: error: ‘numeric_limits’ is not a member of ‘std’
   23 | const double kNegativeInfinity = -std::numeric_limits<double>::infinity();
      |                                        ^~~~~~~~~~~~~~
/home/dkwan/Github/Builds/audio/torchaudio/csrc/decoder/src/decoder/Utils.h:23:55: error: expected primary-expression before ‘double’
   23 | const double kNegativeInfinity = -std::numeric_limits<double>::infinity();

It seems the problem is a missing #include <limits> import in the Utils.h file? I added this in and then I was able to build successfully.

Versions

(This was collected after a successful build after adding the seemingly missing #include line. Also note that my system default compiler is GCC 12, which I did not use as per the command I ran in the previous section)

Collecting environment information...
PyTorch version: 1.13.0a0+gitac8c6d0
Is debug build: False
CUDA used to build PyTorch: 11.6
ROCM used to build PyTorch: N/A

OS: EndeavourOS Linux (x86_64)
GCC version: (GCC) 12.1.0
Clang version: Could not collect
CMake version: version 3.23.2
Libc version: glibc-2.35

Python version: 3.10.4 (main, May 14 2022, 05:21:19) [GCC 12.1.0] (64-bit runtime)
Python platform: Linux-5.18.1-xanmod1-1-native_amd-x86_64-with-glibc2.35
Is CUDA available: True
CUDA runtime version: 11.6.124
GPU models and configuration: GPU 0: NVIDIA GeForce RTX 3090
Nvidia driver version: 515.43.04
cuDNN version: Probably one of the following:
/usr/lib/libcudnn.so.8.4.0
/usr/lib/libcudnn_adv_infer.so.8.4.0
/usr/lib/libcudnn_adv_train.so.8.4.0
/usr/lib/libcudnn_cnn_infer.so.8.4.0
/usr/lib/libcudnn_cnn_train.so.8.4.0
/usr/lib/libcudnn_ops_infer.so.8.4.0
/usr/lib/libcudnn_ops_train.so.8.4.0
HIP runtime version: N/A
MIOpen runtime version: N/A
Is XNNPACK available: True

Versions of relevant libraries:
[pip3] functorch==0.2.0a0+db12967
[pip3] mypy-extensions==0.4.3
[pip3] numpy==1.22.3
[pip3] torch==1.13.0a0+gitac8c6d0
[pip3] torchaudio==0.12.0a0+a63629b
[pip3] torchvision==0.14.0a0+c890a7e
[conda] Could not collect

[mthrok]

Hi @kwand

Thanks for the report. This might be the cause.

@carolineechen Can we add this line? It seems hurmless fix.

[jacobkahn]

Fixed upstream as well in flashlight/text#5 — thanks for reporting, @kwand.