This module generates a JSON compilation database representing how to replay individual compilations for each C++ file in the project. The compilation database is used by Clang Tidy for linting and by some IDEs to offer code completion.
Further reading: JSON Compilation Database Format Specification
Run the following incantation from the project's root directory.
$ ./toolchain/compilation_database/generate_compilation_database.sh
$ python -m json.tool compile_commands.json # Optionally view the formatted JSON
Some build systems offer built-in compilation database generation functionality
such as CMake with its CMAKE_EXPORT_COMPILE_COMMANDS flag. Bazel has no such
feature meaning that Spoor must generate its own.
Spoor's strategy uses a Bazel action listener that
listens to build events and outputs an intermediate Protocol Buffer for each
CppCompile mnemonic. Each intermediate file contains the command to compile a
single source file. Next, we aggregate the intermediate files and inject the
execution directory which was not available during the previous step. Finally,
we format the aggregated result JSON and write the file to
compile_commands.json in Spoor's workspace root.
Although the approach requires regenerating the compilation database on each run, it is more reliable than supporting incremental updates while still demonstrating reasonable performance for this small project.