Linux Package engineering

Creating Linux packages for your Universal Number Library involves several interconnected processes. Lets break down the comprehensive workflow for packaging, building, testing, and distributing your header-only C++ library across major Linux distributions.

Package Format Selection

Since we’re targeting multiple distributions, we’ll need to support different package formats:

• DEB packages (Debian, Ubuntu, derivatives)
• RPM packages (Red Hat, Fedora, SUSE, derivatives)
• AUR packages (Arch Linux)
• Snap/Flatpak (universal, though less common for development libraries)

For a header-only library with CLI utilities, DEB and RPM will be your primary targets.

Build System Integration

Given our modern C++ focus, integrate packaging into the existing build system:

CMake Integration

# Use CPack for cross-platform packaging
include(CPack)
set(CPACK_PACKAGE_NAME "universal-number-lib")
set(CPACK_PACKAGE_VERSION "${PROJECT_VERSION}")
set(CPACK_GENERATOR "DEB;RPM")

# Separate packages per number system
set(CPACK_COMPONENTS_ALL headers cli-utils)
set(CPACK_DEB_COMPONENT_INSTALL ON)
set(CPACK_RPM_COMPONENT_INSTALL ON)

Package Architecture Strategy

For modularity, consider this structure:

• universal-number-lib-dev - Core headers and CMake configs
• universal-number-lib-posit - Posit arithmetic headers
• universal-number-lib-fixpnt - Fixed-point headers
• universal-number-lib-tools - CLI utilities
• Individual packages per number system as needed

Build Process

Automated CI/CD Pipeline

1. GitHub Actions/GitLab CI for automated builds
2. Multi-architecture builds (x86_64, ARM64, potentially RISC-V)
3. Cross-compilation using Docker containers with different base images
4. Package validation using lintian (DEB) and rpmlint (RPM)

Sample workflow structure

- Build and test on Ubuntu LTS, Fedora, Arch
- Generate packages for each architecture
- Run package installation tests
- Upload to staging repositories

Quality Assurance Process

Pre-release QA

• Static analysis integration (clang-tidy, PVS-Studio)
• Package linting (lintian, rpmlint, namcap)
• Installation testing on clean VMs/containers
• Reverse dependency testing if other packages depend on yours
• ABI compatibility checking using abidiff tools

Testing matrix

• Multiple compiler versions (GCC 9-13, Clang 10-17)
• Different C++ standard versions (C++17, C++20, C++23)
• Various optimization levels and sanitizer builds

Distribution Registration Process

Debian/Ubuntu

1. Mentors process - Upload to mentors.debian.net for sponsor review
2. ITP bug filing (Intent To Package) in Debian BTS
3. Packaging standards compliance (Debian Policy, debhelper compat levels)
4. Sponsor finding for initial uploads until you become a Debian Developer

Fedora

1. Package Review Process - Submit spec file for community review
2. Fedora Package Database registration
3. Bugzilla integration for package maintenance
4. Koji build system integration

Arch Linux

1. AUR submission - Create PKGBUILD and submit to AUR
2. Community maintenance or eventual migration to official repositories

Release Management

Versioning Strategy

• Semantic versioning aligned with library releases
• Package revision numbers for packaging-only changes
• Epoch handling for version scheme changes

Release Automation

# Automated release pipeline

1. Tag release in Git

2. Trigger CI builds for all targets

3. Generate changelogs from Git history

4. Upload packages to staging

5. Run integration tests

6. Promote to production repositories

Specialized Considerations for Header-Only Libraries

Package Content

• Install headers to /usr/include/universal/
• CMake config files in /usr/lib/cmake/universal/
• pkg-config files for non-CMake builds
• Documentation and examples

Development Experience

# Users should be able to simply:
find_package(Universal COMPONENTS posit fixpnt REQUIRED)

target_link_libraries(myapp Universal::posit Universal::fixpnt)

Repository Hosting Options

Personal Package Archives (PPAs)

• Launchpad PPAs for Ubuntu
• Copr for Fedora/RHEL
• OBS (Open Build Service) for cross-distribution builds

Corporate/Professional

• JFrog Artifactory
• Sonatype Nexus
• CloudSmith for cloud-native package management

Long-term Maintenance Strategy

Automated maintenance

• Dependabot/Renovate for dependency updates
• Automated security scanning
• License compliance checking
• Documentation generation and deployment

Community engagement

• Clear contribution guidelines
• Issue templates for packaging problems
• Distribution-specific maintainer relationships

The complexity scales with the ambition to provide an easy to use mixed-precision algorithm development and optimization SDK. Starting with PPA/Copr for testing, then moving to official repositories as adoption grows. Given the specialized nature of numerical computing libraries, consider also targeting HPC-focused distributions and package managers like Spack, which are popular in scientific computing communities where this library should see heavy use.