The purpose of this project is to demonstrate how to include flatcc as dependency of a small C program that builds flatbuffers. This works by using the Meson build system which is a more userfriendly CMake/Ninja style build system with a sort of built-in package management.
This project depends on the meson branch of flatcc which is currently experimental.
We use a the user provided test case for flatcc issue 4 as an example.
To get started, please inspect the subprojects/flatcc.wrap file to
ensure the git url and revision is up to date.
Dependencies: C compiler tool chain, git, ninja-build, python3, and meson 0.41.0+.
The setup-build.sh discussed below will fetch meson if not installed.
Note: By design Meson only depends on core python3 libraries and can be run from source checked out via github.
To fetch, build and run test:
step 1: mkdir build
step 2: meson build (or meson.py build)
step 3: ninja -C build test
step 4: there is no step 4
or you can call setup-build.sh which does the same.
Yes, now you can have automagically built schema headers with solid build dependencies with very little setup.
To test dependencies, try
$ ninja -C build
nothing to build
$ thouch schema/*.fbs
$ ninja -C build
linking app
Note that this also works with schema files that include other schema
files as is the case in the main flatcc monster_test project, but here
we have a lighter setup. The only problem is ninja -C build clean
might complain a bit because of unknown output products, But these are
always created via solid dependencies unless the build is manipulated
directly.
Note that Meson can choose a system installed version of flatcc over fetching a subproject with git. However, because we some nice complation rules in the subproject, and because it is really fast to fetch and compile, it is much better to embed flatcc as subproject always. Just make sure to link executable static. Note: linking statically typically results in binaries much smaller that the static library or the shared library size alone, like a factor 10.
If working with many flatbuffers projecst, it is also possible to download the flatcc repo once and symbolically link it into subjprojects to avoid repeated downloading.
Also notice the .gitignore file: we ignore our build directory and everything in the subproject dir except for the .wrap file that fetch the flatcc repository. The repository is placed in subprojects by git, but is ignored by our parent git repo. The flatcc project is built in the build dir and does not touch subprojects after being fetched.
Build times on 4 core i7, 2,2GHz, Ninja/Clang/MacOS
To set up a new project after cloning this repo, but without having fetched flatcc yet:
$ time ./setup-build.sh
4.10 real 5.53 user 1.29 sys
The above primarily measures the time it takes to clone the flatcc repo via meson.
Build times broken down after the initial clone of flatcc.
$ rm -rf build/*
$ time meson build
0.23 real 0.16 user 0.06 sys
$ time ninja -C build test
1.29 real 5.15 user 1.10 sys
# (again)
$ time ninja -C build test
0.17 real 0.13 user 0.03 sys
# (again without test)
$ time ninja -C build
ninja: Entering directory `build'
ninja: no work to do.
0.00 real 0.00 user 0.00 sys
$ touch schema/*.fbs
$ time ninja -C build
ninja: Entering directory `build'
[3/3] Linking target flatcc-issue4
0.11 real 0.08 user 0.02 sys