You must have these available on the host system:
- Python 3.7+
- Docker
To run CUDA tests, you will also need:
- NVIDIA Docker runtime
- CUDA Driver 396.26+
(These are just the requirements for the "host system" that runs the docker containers. ISCE3 itself has a much more extensive set of prerequisites, which will be automatically installed inside the docker containers.)
./tools/run.py allThis will run the typical "all" sequence which is intended to mimic our CI build/testing pipeline. More options can be specified, e.g.
./tools/run.py -B build-alpine -i alpine setupThis will run the "setup" step for the "alpine" image set, using a build directory named "build-alpine". See the run.py script for a list of steps, and imagesets/imgset.py for their default implementations.
CI failures should be reproducible locally via these scripts. Running the "all" sequence should be sufficient.
./tools/run.py allIf additional info is needed, you can "drop in" to the builder image and run commands interactively. E.g.
./tools/run.py dropinThis will put you in the build directory with the same environment the tests ran in. You can then run commands to isolate the error:
ctest -R my_failing_test --verboseThese scripts can be used as a local development environment, to sidestep installing all the development libraries required to build ISCE3. First build the docker images, run CMake configuration, and compile the code.
./tools/run.py setup configure buildAfter some local development, you can just rerun the build step - CMake will only recompile files which have been modified.
./tools/run.py buildSee Building identical conda environments
- Create a fresh conda environment
- Install runtime dependencies from requirements.txt
conda list --explicit > runtime/spec-file.txt- Install dev dependencies from requirements.txt
conda list --explicit > dev/spec-file.txt- Make sure the dev dependencies are a superset of the runtime dependencies