[Build] Build failed when --use_tensorrt

[FredYakumo]

Describe the issue

System: Arch Linux
--use_cuda
Build comand is, and the related path are right at all:

./build.sh --config Release --update --build --use_cuda --use_tensorrt --cuda_home=/usr/local/cuda/ --cudnn_home=/usr --tensorrt_home=/usr/local/TensorRT/  --use_tensorrt_oss_parser --parallel --cmake_generator Ninja

Get error:
NVCC_OUT = 1
-- CMAKE_CUDA_COMPILER_VERSION: 12.9.41
-- Enable flash attention for CUDA EP
-- Enable lean attention for CUDA EP
-- Enable memory efficient attention for CUDA EP
-- CMAKE_CXX_COMPILER_ID: GNU
-- CMAKE_CXX_COMPILER_VERSION: 11.4.0
-- Using -mavx2 -mfma -mavxvnni flags
CMake Warning (dev) at /usr/local/share/cmake-4.0/Modules/FetchContent.cmake:1953 (message):
Calling FetchContent_Populate(cutlass) is deprecated, call
FetchContent_MakeAvailable(cutlass) instead. Policy CMP0169 can be set to
OLD to allow FetchContent_Populate(cutlass) to be called directly for now,
but the ability to call it with declared details will be removed completely
in a future version.
Call Stack (most recent call first):
external/cutlass.cmake:11 (FetchContent_Populate)
onnxruntime_providers_cuda.cmake:256 (include)
onnxruntime_providers_cuda.cmake:320 (config_cuda_provider_shared_module)
onnxruntime_providers.cmake:124 (include)
CMakeLists.txt:1865 (include)
This warning is for project developers. Use -Wno-dev to suppress it.

CMake Error at onnxruntime_providers_tensorrt.cmake:42 (math):
math cannot parse the expression: "": syntax error, unexpected end of file
(0).
Call Stack (most recent call first):
onnxruntime_providers.cmake:132 (include)
CMakeLists.txt:1865 (include)

CMake Error at onnxruntime_providers_tensorrt.cmake:43 (math):
math cannot parse the expression: "": syntax error, unexpected end of file
(0).
Call Stack (most recent call first):
onnxruntime_providers.cmake:132 (include)
CMakeLists.txt:1865 (include)

CMake Error at onnxruntime_providers_tensorrt.cmake:44 (math):
math cannot parse the expression: "": syntax error, unexpected end of file
(0).
Call Stack (most recent call first):
onnxruntime_providers.cmake:132 (include)

CMakeLists.txt:1865 (include)

-- Can't find NV_TENSORRT_MAJOR macro
-- Looking for nvinfer and nvinfer_plugin
-- Found CUDA headers at /usr/local/cuda/targets/x86_64-linux/include
-- Found TensorRT headers at /usr/local/TensorRT/include

Troubleshooting and Fix:
After exploring various build option adjustments, I discovered that the root cause was an improperly matched regex pattern. By modifying the regex string match in onnxruntime_providers_tensorrt.cmake, I resolved the parsing issue and achieved a successful build.

You can review my proposed fix here:6a1b4e6

Urgency

No response

Target platform

Arch Linux on WSL2

Build script

./build.sh --config Release --update --build --use_cuda --use_tensorrt --cuda_home=/usr/local/cuda/ --cudnn_home=/usr --tensorrt_home=/usr/local/TensorRT/ --use_tensorrt_oss_parser --parallel --cmake_generator Ninja

Error / output

NVCC_OUT = 1
-- CMAKE_CUDA_COMPILER_VERSION: 12.9.41
-- Enable flash attention for CUDA EP
-- Enable lean attention for CUDA EP
-- Enable memory efficient attention for CUDA EP
-- CMAKE_CXX_COMPILER_ID: GNU
-- CMAKE_CXX_COMPILER_VERSION: 11.4.0
-- Using -mavx2 -mfma -mavxvnni flags
CMake Warning (dev) at /usr/local/share/cmake-4.0/Modules/FetchContent.cmake:1953 (message):
Calling FetchContent_Populate(cutlass) is deprecated, call
FetchContent_MakeAvailable(cutlass) instead. Policy CMP0169 can be set to
OLD to allow FetchContent_Populate(cutlass) to be called directly for now,
but the ability to call it with declared details will be removed completely
in a future version.
Call Stack (most recent call first):
external/cutlass.cmake:11 (FetchContent_Populate)
onnxruntime_providers_cuda.cmake:256 (include)
onnxruntime_providers_cuda.cmake:320 (config_cuda_provider_shared_module)
onnxruntime_providers.cmake:124 (include)
CMakeLists.txt:1865 (include)
This warning is for project developers. Use -Wno-dev to suppress it.

CMake Error at onnxruntime_providers_tensorrt.cmake:42 (math):
math cannot parse the expression: "": syntax error, unexpected end of file
(0).
Call Stack (most recent call first):
onnxruntime_providers.cmake:132 (include)
CMakeLists.txt:1865 (include)

CMake Error at onnxruntime_providers_tensorrt.cmake:43 (math):
math cannot parse the expression: "": syntax error, unexpected end of file
(0).
Call Stack (most recent call first):
onnxruntime_providers.cmake:132 (include)
CMakeLists.txt:1865 (include)

CMake Error at onnxruntime_providers_tensorrt.cmake:44 (math):
math cannot parse the expression: "": syntax error, unexpected end of file
(0).
Call Stack (most recent call first):
onnxruntime_providers.cmake:132 (include)

CMakeLists.txt:1865 (include)

-- Can't find NV_TENSORRT_MAJOR macro
-- Looking for nvinfer and nvinfer_plugin
-- Found CUDA headers at /usr/local/cuda/targets/x86_64-linux/include
-- Found TensorRT headers at /usr/local/TensorRT/include

Visual Studio Version

No response

GCC / Compiler Version

No response