Skip to content

Packaging.md

Latest commit

 

History

History
73 lines (37 loc) · 2.51 KB

Packaging.md

File metadata and controls

73 lines (37 loc) · 2.51 KB

Packaging for Tiny Tapeout

Instructions to rebuild the Docker container & Python wheels used in the Tiny Tapeout GitHub actions.

1. Build & test the nix package

Run nix-shell. This will rebuild if necessary, then drop you into a shell with the openlane2 and its dependencies available via nix. Exit the shell when done testing.

2. Build the docker container

IMAGE_PATH=$(nix build --extra-experimental-features "flakes nix-command" --print-out-paths --no-link --accept-flake-config --option system x86_64-linux --extra-platforms x86_64-linux '.#packages.x86_64-linux.openlane-docker')

cat $IMAGE_PATH | docker load

3. Build wheels for libparse-python

(libparse-python doesn't change often, so you might reuse the wheels from the last TinyTapeout/libparse-python release.)

Clone https://github.com/TinyTapeout/libparse-python and change into the repo directory.

Build & install for the current python version:

  • pip install .

Use cibuildwheel to build a manylinux wheel:

  • pip install cibuildwheel, if necessary

  • CIBW_SKIP="pp*" python -m cibuildwheel --output-dir wheelhouse

Built wheels are in the wheelhouse directory.

4. Build wheel for openlane2

Return to the openlane2 directory.

Update the version in pyproject.toml.

Build & install openlane2 for the current python version:

  • pip install .

Make a redistributable wheel:

  • Make a wheelhouse subdirectory and copy the contents of the libparse-python/wheelhouse (or the wheels downloaded from the github release) into it.

  • pip wheel -w wheelhouse -f wheelhouse .

5. Test it

Run OPENLANE_IMAGE_OVERRIDE=openlane:tmp-x86_64-linux openlane --dockerized.

6. Upload the wheels

If libparse-python was changed, create a new tag & relase under TinyTapeout/libparse-python and upload all the wheels in the libparse-python/wheelhouse directory.

Create a new tag & relase under TinyTapeout/openlane2. Upload the newly created openlane-<version>.whl file from openlane2/wheelhouse. There will be other wheels for the dependencies (and possibly older versions of openlane2 built earlier), ignore those.

7. Upload the docker container

docker tag openlane:tmp-x86_64-linux ghcr.io/tinytapeout/openlane2:<version>

docker push ghcr.io/tinytapeout/openlane2:<version>

8. Update the action

Edit action.yml in TinyTapeout/tt-gds-action to change:

  • Container version in OPENLANE_IMAGE_OVERRIDE

  • Wheel versions in pip install <whatever>.whl lines

Check if an example project hardens successfully.