Compilation Database Generation
To analyze a complete project, you need a compilation database containing all
the compiler instructions for its source files. In this guide, we walk you through
generating a compile_commands.json file , demonstrated with a Fortran open
source project.
Getting ready
For this demonstration, we will use the Fortran implementation of the Himeno benchmark, a Poisson equation solver. Start by cloning the repository:
git clone https://github.com/codee-com/codee-demos.git
Navigate to the source code directory:
cd codee-demos/Fortran/Himeno
Third party tools to generate the compile_commands.json
To analyze a complete project with many source files, you need a compilation database, which is
a file typically named compile_commands.json containing all the compiler invocations for the
source files. It provides Codee with information about the name, version and location of the
compiler in the system, required compiler flags, and location of the target source files.
For Linux, the recommendation is to use Bear.
For Windows, the recommendation is to use CMake with Ninja generators.
| Linux | Compilers | Compiler wrappers | Windows | |
|---|---|---|---|---|
| bear | ✔ | ✔ | ✔ | ✘ |
| CMake | ✔ | ✔ | codee CLI | Ninja generator |
Use bear (Recommended)
Ensure you are using the bear tool distributed with Codee, not other
installations of bear, in order to take advantage of its enhanced
support for compilers (e.g. GNU, LLVM, Intel, AMD, Nvidia, Cray) and compiler
wrappers (e.g. MPI wrappers in supercomputers, Microsoft wrappers).
The recommended way of generating the compilation database is to use
bear, which comes bundled with
Codee packages. bear is able to intercept the compiler calls made by
your build system in order to generate the compile_commands.json file.
To use it you only need to prepend bear -- to your typical build command.
To generate it for the Himeno project copy the following command:
bear -- make
Since Himeno is a single-file project, the compilation database will have a single entry:
[
{
"arguments": [
"/usr/bin/gfortran",
"-c",
"-O3",
"-I",
"/usr/include",
"-I",
".",
"-o",
"himeno",
"himeno.f90"
],
"directory": "/home/user/codee-demos/Fortran/Himeno",
"file": "/home/user/codee-demos/Fortran/Himeno/himeno.f90",
"output": "/home/user/codee-demos/Fortran/Himeno/himeno"
}
]
Running codee commands with the additional --db codee.db flag enables
Incremental Static Analysis. This reduces runtime by storing analysis
results and reusing them in subsequent analysis, reanalyzing only
the source code that has changed.
Now you can use the JSON generated to substitute the compilation command in the Codee invocation. For example:
codee checks --check-id PWR069 -p compile_commands.json --db codee.db
Date: 2026-03-25 Codee version: 2026.1 License type: Team
Searching Incremental Static Analysis database... Enabled
[1/1] himeno.f90 ... Done
QUALITY CHECKS REPORT
himeno.f90:137:1 [PWR069] (level: L2): Use the keyword only to explicitly state what to import from a module
himeno.f90:165:1 [PWR069] (level: L2): Use the keyword only to explicitly state what to import from a module
himeno.f90:224:1 [PWR069] (level: L2): Use the keyword only to explicitly state what to import from a module
himeno.f90:256:1 [PWR069] (level: L2): Use the keyword only to explicitly state what to import from a module
himeno.f90:276:1 [PWR069] (level: L2): Use the keyword only to explicitly state what to import from a module
OPTIMIZATION CHECKS REPORT
No actionable items were found
SUGGESTIONS
Use --check-id and --verbose to focus on specific subsets of checkers, e.g.:
codee checks --check-id PWR069 --verbose -p compile_commands.json --db codee.db
1 built file (0 results taken from cache), 0 dependencies (0 reused from cache)
1 target file, 7 functions, 5 loops, 214 SLOCs successfully analyzed (5 checkers) and 0 non-analyzed files in 351 ms
Using CMake (Alternative)
Ensure you are using CMake version 3.5 or higher.
As an alternative, CMake has its own way of obtaining the compile_commands.json file.
Add -DCMAKE_EXPORT_COMPILE_COMMANDS=ON to your usual CMake invocation; for instance:
cmake . -DCMAKE_Fortran_COMPILER=gfortran -DCMAKE_EXPORT_COMPILE_COMMANDS=ON -G "Ninja" -B build
The compile_commands.json file will be now stored in the build/ directory and
it will look like this:
[
{
"directory": "/home/user/codee-demos/Fortran/Himeno/build",
"command": "/usr/bin/gfortran -c /home/user/codee-demos/Fortran/Himeno/himeno.f90 -o CMakeFiles/himeno.dir/himeno.f90.o",
"file": "/home/user/codee-demos/Fortran/Himeno/himeno.f90",
"output": "CMakeFiles/himeno.dir/himeno.f90.o"
}
]
Now you can use the JSON generated to substitute the compilation command in the Codee invocation. For example:
codee checks --check-id PWR069 -p build/compile_commands.json --db codee.db
Date: 2026-03-25 Codee version: 2026.1 License type: Team
Searching Incremental Static Analysis database... Enabled
[1/1] /home/user/codee-demos/Fortran/Himeno/himeno.f90 ... Done
QUALITY CHECKS REPORT
/home/user/codee-demos/Fortran/Himeno/himeno.f90:137:1 [PWR069] (level: L2): Use the keyword only to explicitly state what to import from a module
/home/user/codee-demos/Fortran/Himeno/himeno.f90:165:1 [PWR069] (level: L2): Use the keyword only to explicitly state what to import from a module
/home/user/codee-demos/Fortran/Himeno/himeno.f90:224:1 [PWR069] (level: L2): Use the keyword only to explicitly state what to import from a module
/home/user/codee-demos/Fortran/Himeno/himeno.f90:256:1 [PWR069] (level: L2): Use the keyword only to explicitly state what to import from a module
/home/user/codee-demos/Fortran/Himeno/himeno.f90:276:1 [PWR069] (level: L2): Use the keyword only to explicitly state what to import from a module
OPTIMIZATION CHECKS REPORT
No actionable items were found
SUGGESTIONS
Use --check-id and --verbose to focus on specific subsets of checkers, e.g.:
codee checks --check-id PWR069 --verbose -p build/compile_commands.json --db codee.db
1 built file (0 results taken from cache), 0 dependencies (0 reused from cache)
1 target file, 7 functions, 5 loops, 214 SLOCs successfully analyzed (5 checkers) and 0 non-analyzed files in 337 ms