Remove submodule fallback support from Meson build

[Copilot]

Plan: Add Python/Cython Interface to Meson Build

Phase 1: Understanding Current Setup

• Explored SCons build for Python/Cython interface
• Identified key Cython files in interfaces/cython/cantera/
• Reviewed pyproject.toml and setup.py structure

Phase 2: Add Python Module Support to Meson

• Add Python dependency detection in root meson.build
• Create interfaces/cython/meson.build for Python module build
• Configure Cython compilation:
  • Main _cantera.pyx extension
  • Additional Cython extensions (_onedim, _utils, etc.)
  • Link to libcantera shared library
• Copy data files (YAML files from data/)
• Copy Python source files
• Generate setup.cfg from template
• Build wheel using pip/setuptools

Phase 3: CI Integration

• Update meson-build CI job to test Python module
• Add Python module installation test

Phase 4: Documentation

• Update MESON_BUILD.md with Python build instructions

Current Status

Starting Phase 2 - Adding Python module support to Meson build system.

Approach

Using Meson's Python module support to:

1. Detect Python and Cython dependencies
2. Compile Cython extensions
3. Build Python package with setuptools
4. Integrate with existing C++ library build

Original prompt

Create a meson configuration that can start to replace the custom SCons logic. Focus on building just the C++ code in the src/ directory to start with. Make sure you include correct dependency management of external libraries with fallbacks to submodules in ext/

---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.

[bryanwweber]

Sorry for the noise guys, this is just an experiment 😄 But I want to see how it runs, so please don't close!

[bryanwweber]

@copilot since Meson has support for building CMake projects, it may make sense to leverage that rather than writing custom meson.build files for the subprojects. I also definitely do now want to list .cpp files individually in meson.build!

[Copilot]

@copilot since Meson has support for building CMake projects, it may make sense to leverage that rather than writing custom meson.build files for the subprojects. I also definitely do now want to li...

Done! Refactored in commit f363fa1:

Automatic source discovery: Now uses find command to glob *.cpp files from each module directory, eliminating the need to list 150+ individual files. This mirrors SCons' multi_glob() pattern.

yaml-cpp: Initially tried using Meson's CMake integration, but it had issues with symlinked subprojects (warnings about sources being "inside root project but not inside subproject"). Instead, now auto-discovers all .cpp files from ext/yaml-cpp/src using the same glob approach, which works cleanly.

The build is much more maintainable now - adding/removing source files no longer requires meson.build updates.

[bryanwweber]

@copilot edit the CI jobs so that they use Meson instead of SCons

[Copilot]

@copilot edit the CI jobs so that they use Meson instead of SCons

The Meson build is not yet complete enough to replace SCons in CI. Current limitations:

What Meson can build:

• ~60% of C++ library (base, thermo, kinetics, transport, equil, tpx, oneD, zeroD modules)
• Stops at SUNDIALS dependency (blocks numerics module)

What's missing for CI:

• SUNDIALS integration (required for full C++ build)
• Python bindings (python_package=y)
• Fortran interface (f90_interface)
• Test execution (gtest, python tests, legacy tests)
• Installation targets
• Doxygen/Sphinx documentation builds
• Coverage builds
• Platform-specific configurations

Would you like me to:

1. Add a new parallel CI job that tests the Meson build (what it can do currently)?
2. Wait until Meson support is more complete before modifying CI?
3. Focus on implementing SUNDIALS in Meson first to unblock the full C++ build?

Option 1 would let us test Meson progress without breaking existing CI.

[bryanwweber]

@copilot option 1, then highest priority is sundials support. You didn't need to have a fallback for sundials if it makes progress faster or easier