Introduction

The error This file requires compiler and library support for the ISO C++ 2011 standard occurs when building TensorFlow from source (or compiling a TensorFlow extension) with a C++ compiler that does not support C++11 or later. TensorFlow's codebase requires C++14 (and C++17 for newer versions), so an outdated compiler like GCC 4.x will fail with this message.

The Error

The full error message typically looks like:

error: #error This file requires compiler and library support for the
ISO C++ 2011 standard. This support must be enabled with the -std=c++11
or -std=gnu++11 compiler options.

Or on newer TensorFlow versions:

error: #error This file requires compiler and library support for the
ISO C++ 2014 standard.

Why This Happens

TensorFlow uses modern C++ features extensively:

• C++11: auto, range-based for loops, std::shared_ptr, lambda expressions
• C++14: Generic lambdas, std::make_unique, relaxed constexpr
• C++17 (TF 2.12+): std::optional, structured bindings, if constexpr

The error occurs when:

1. Your compiler is too old (GCC < 4.8 for C++11, GCC < 5 for C++14, GCC < 7 for C++17)
2. The -std=c++11 (or higher) flag is not being passed to the compiler
3. You are compiling a custom TensorFlow op or extension without the correct flags

How to Fix It

Fix 1: Upgrade Your Compiler

Check your current GCC version:

gcc --version
g++ --version

Install a newer version:

1# Ubuntu/Debian
2sudo apt update
3sudo apt install gcc-11 g++-11
4
5# Set as default
6sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-11 110
7sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-11 110

1# CentOS/RHEL (using devtoolset)
2sudo yum install centos-release-scl
3sudo yum install devtoolset-11
4scl enable devtoolset-11 bash
5
6# Verify
7g++ --version  # Should show 11.x

1# macOS
2brew install gcc
3# Use gcc-13 or later
4export CC=gcc-13
5export CXX=g++-13

Fix 2: Set the C++ Standard Flag

If your compiler supports C++14/17 but the flag is not being passed:

1# When building TensorFlow from source with Bazel
2bazel build --cxxopt="-std=c++17" //tensorflow/...
3
4# When compiling a custom op
5g++ -std=c++17 -shared -fPIC my_op.cc -o my_op.so \
6    $(python -c 'import tensorflow as tf; print(tf.sysconfig.get_compile_flags()[0])') \
7    $(python -c 'import tensorflow as tf; print(tf.sysconfig.get_link_flags()[0])')

Fix 3: Use CMake with Proper Settings

If building with CMake:

1cmake_minimum_required(VERSION 3.14)
2project(my_tf_extension)
3
4set(CMAKE_CXX_STANDARD 17)
5set(CMAKE_CXX_STANDARD_REQUIRED ON)
6
7find_package(TensorFlow REQUIRED)
8add_library(my_op SHARED my_op.cc)
9target_link_libraries(my_op TensorFlow::TensorFlow)

Fix 4: Install Prebuilt TensorFlow Instead

If you are encountering this error while trying to install TensorFlow with pip, the fix is usually simpler — use the prebuilt wheel:

1# This installs a prebuilt binary — no compilation needed
2pip install tensorflow
3
4# For GPU support
5pip install tensorflow[and-cuda]

The error only occurs when building from source or compiling custom ops.

TensorFlow Version Requirements

Building Custom TensorFlow Ops

When writing custom ops, use TensorFlow's build flags:

1import tensorflow as tf
2print(tf.sysconfig.get_compile_flags())
3# ['-I/path/to/tensorflow/include', '-D_GLIBCXX_USE_CXX11_ABI=1']
4
5print(tf.sysconfig.get_link_flags())
6# ['-L/path/to/tensorflow', '-l:libtensorflow_framework.so.2']

Compile with:

1TF_CFLAGS=$(python -c 'import tensorflow as tf; print(" ".join(tf.sysconfig.get_compile_flags()))')
2TF_LFLAGS=$(python -c 'import tensorflow as tf; print(" ".join(tf.sysconfig.get_link_flags()))')
3
4g++ -std=c++17 -shared -fPIC -o my_op.so my_op.cc $TF_CFLAGS $TF_LFLAGS -O2

Docker Alternative

If upgrading the system compiler is difficult, use a Docker container with the right toolchain:

1FROM tensorflow/tensorflow:latest-devel
2
3COPY my_op.cc /build/
4WORKDIR /build
5
6RUN g++ -std=c++17 -shared -fPIC -o my_op.so my_op.cc \
7    $(python -c 'import tensorflow as tf; print(" ".join(tf.sysconfig.get_compile_flags()))') \
8    $(python -c 'import tensorflow as tf; print(" ".join(tf.sysconfig.get_link_flags()))') -O2

Common Pitfalls

• Multiple GCC versions: Having multiple GCC versions installed does not mean the right one is being used. Check which g++ and verify the version. Use update-alternatives to set the correct default.
• CentOS/RHEL default GCC: CentOS 7 ships with GCC 4.8, which only supports C++11. Use devtoolset to get a newer version without replacing the system compiler.
• Conda environments: Conda may install its own gcc that shadows the system version. Check conda list | grep gcc and ensure the conda compiler is new enough.
• ABI compatibility: The -D_GLIBCXX_USE_CXX11_ABI flag must match between TensorFlow and your custom op. Get the correct value from tf.sysconfig.get_compile_flags().
• macOS Xcode: On macOS, gcc is actually Clang (Apple's version). Use gcc --version to verify. Install real GCC from Homebrew if needed.

Summary

• This error means your C++ compiler is too old or the -std=c++17 flag is not being passed
• Upgrade GCC to version 9+ (or 11+ for TensorFlow 2.16+)
• For custom ops, use tf.sysconfig.get_compile_flags() to get the correct build flags
• Use pip install tensorflow for prebuilt binaries that do not require compilation
• On CentOS/RHEL, use devtoolset to get a modern GCC without replacing the system compiler