Boost Mobile Crash Debugging with Symbolic Demangle: A Complete Walkthrough

1. Introduction

Huolala's mobile platform has a comprehensive stability monitoring system that helps create industry‑leading products, issue alerts, and root‑cause analysis. This article shares the practice of the symbolic‑demangle direction and opens the demangle tool on GitHub for others.

2. What is symbolic demangle?

In programming, demangling restores compiler‑mangled names to readable symbols. Mangling (name mangling) encodes function and variable names with type information, especially in C++ where overloads and namespaces exist. Demangling is essential for debugging, stack tracing, and crash diagnosis across languages like C++, Rust, and Swift.

The following image shows an undecoded stack trace with poor readability:

3. Why choose the symbolic solution?

Huolala's mobile stack mainly uses C++ and Swift. Existing tools (c++filt, swift‑demangle) require language detection, environment setup, and efficiency trade‑offs. An article from ByteDance highlighted the use of the Rust‑implemented symbolic‑demangle to replace LLVM's demangler, reducing operational cost. After evaluating open‑source options, symbolic‑demangle was selected for its low cost and effectiveness.

4. Symbolic practice in Huolala

Symbolic currently supports Swift up to 5.5.1; higher versions need adaptation. The following sections describe adapting to newer Swift versions and building a Python library on Linux.

4.1 Adapting to newer Swift

The adaptation involves two steps: syncing the official Swift demangle source and compiling the symbolic project.

Sync Swift source:

mkdir swift-source && cd swift-source

git clone [email protected]:swiftlang/swift.git

./swift/utils/update-checkout --clone

cd swift

git checkout swift-5.10-RELEASE

./utils/build-script --skip-build

After updating sources, copy them to the library: ./update.py swift-source Compile the symbolic Rust project (uses Cargo): cargo build Common build issues and fixes:

Missing LLVM files – copy from llvm‑project.

Missing template names – update type_traits.h from Swift source.

Standard library errors – compile with C++17.

Stale build artifacts – delete target directory.

4.2 Building the Python library on Linux

Generate a wheel package: python3 setup.py bdist_wheel Install the wheel:

pip3 install dist/symbolic-10.2.1-py2.py3-none-macosx_14_0_universal2.whl

Linux builds encountered GLIBC version mismatches (required 2.29, Debian 12 provides 2.36). Solutions include using a lower‑version OS (CentOS 7 with GLIBC 2.17) or adjusting GCC versions.

On CentOS 7, upgrade GCC to support C++17 and install required dependencies:

yum -y install centos-release-scl

yum clean all

yum makecache

yum -y install libffi-devel python3-devel

After resolving compiler and library issues, the wheel builds successfully.

5. Final results

Online demangling output matches Xcode's swift‑demangle tool closely.

Local symbolic demangle output:

Xcode swift‑demangle output for comparison:

6. References

Name mangling – Wikipedia

iOS Symbol Parsing Refactoring

symbolic‑demangle GitHub repository