Skip to content

BUILD_LINUX.md

Latest commit

 

History

History
153 lines (119 loc) · 4.24 KB

BUILD_LINUX.md

File metadata and controls

153 lines (119 loc) · 4.24 KB

LoliProfiler Linux Build Guide

Overview

This branch contains fixes for building LoliProfiler CLI on Linux platforms. The build system has been enhanced to support CLI-only builds without requiring GUI dependencies.

Prerequisites

Required Packages

For CentOS/RHEL:

yum install -y qt5-qtbase-devel cmake gcc-c++ make zip

For Ubuntu/Debian:

apt-get install -y qtbase5-dev qt5-qmake cmake g++ make zip

Building

One-Click Build

Simply run the build script:

./build_linux.sh

This script will:

  1. Configure CMake with CLI-only build option (-DBUILD_GUI=OFF)
  2. Build the LoliProfilerCLI binary
  3. Deploy Qt dependencies automatically
  4. Copy Python analysis scripts
  5. Create a distribution package: dist/LoliProfiler-linux-cli.zip

Manual Build

If you prefer to build manually:

# Create build directory
mkdir -p build/cmake
cd build/cmake

# Configure (CLI only)
cmake ../.. \
    -DCMAKE_BUILD_TYPE=Release \
    -DCMAKE_PREFIX_PATH=/path/to/qt5 \
    -DBUILD_GUI=OFF

# Build
make -j$(nproc) LoliProfilerCLI

# Install
make install

Running

After building, the distribution package will be in dist/ directory. Extract and run:

cd dist/LoliProfiler
./LoliProfilerCLI.sh --help

The launcher script (LoliProfilerCLI.sh) automatically sets the correct library paths.

Changes Made

CMakeLists.txt

  • Added BUILD_GUI option (default: ON for backward compatibility)
  • GUI application build is now conditional based on BUILD_GUI flag
  • Linux builds can use -DBUILD_GUI=OFF to build CLI only

build_linux.sh

  • Complete rewrite based on build.bat (Windows build script)
  • Auto-detects Qt5 installation
  • Builds CLI-only by default
  • Calls deployment script to copy dependencies
  • Creates final ZIP distribution package

scripts/Deployqt_linux.sh

  • New script equivalent to Windows windeployqt
  • Automatically detects and copies Qt dependencies
  • Creates launcher script with proper LD_LIBRARY_PATH
  • Copies Qt plugins (platforms, xcbglintegrations)

Code Fixes

Fixed Qt 5.15+ compatibility issues:

  • Replaced deprecated QString::SkipEmptyParts with Qt::SkipEmptyParts
  • Replaced deprecated endl with Qt::endl
  • Added missing #include <QPointF> in profilecomparator.cpp
  • Fixed unused parameter warnings in NO_GUI_MODE

Distribution Package Contents

LoliProfiler/
├── LoliProfilerCLI          # Main CLI binary
├── LoliProfilerCLI.sh       # Launcher script (use this to run)
├── qt.conf                  # Qt configuration
├── lib/                     # Qt and other dependencies
├── plugins/                 # Qt plugins
├── analyze_heap.py          # MCP-based AI analysis script
├── markdown_to_html.py      # Markdown to HTML converter
├── requirements.txt         # Python dependencies
├── mcp_server/              # MCP heap explorer server
│   ├── __init__.py
│   ├── tree_model.py
│   └── heap_explorer_server.py
└── .mcp.json                # MCP server configuration

Platform Compatibility

The changes are designed to:

  • ✅ Not affect Windows builds (BUILD_GUI defaults to ON)
  • ✅ Not affect macOS builds (BUILD_GUI defaults to ON)
  • ✅ Enable Linux CLI-only builds with minimal CMake changes
  • ✅ Use platform-specific #ifdef blocks where needed

Usage Examples

Profile an Android app

./LoliProfilerCLI.sh --app com.example.app --out profile.loli --duration 60

Compare two profiles

./LoliProfilerCLI.sh --compare baseline.loli comparison.loli

Attach to running app

./LoliProfilerCLI.sh --app com.example.app --attach --out profile.loli

Troubleshooting

Qt not found

If CMake cannot find Qt5, set the QT5Path environment variable:

export QT5Path=/usr/lib64/qt5  # or your Qt5 installation path
./build_linux.sh

Missing libraries at runtime

Use the provided launcher script LoliProfilerCLI.sh instead of running the binary directly.

Notes

  • GUI modules are NOT built on Linux by default (saves compilation time and dependencies)
  • All platform-specific code is wrapped in #ifdef blocks to ensure cross-platform compatibility
  • The deployment script mimics Windows windeployqt behavior for consistency