GitLab

Fix build

Fix build

Fix build

Fix build

Fix Makefile

Fix build

Fix build

Fix build

Fix build

Add back conda build, update numpy dep

Fix build

Fix build

Fix build

Fix build

Fix build

Fix build

In principle, now just running `conda env create -f environment.yml -n finesse3` is enough to
get Finesse 3 up and running, including the running of setup. On my machine I can run `pytest`
immediately after the above command and it passes straight away.

This is an attempt to use pyproject.toml for as much project configuration as possible. Only
flake8 doesn't use it, and instead a .flake8 config file is created. The idea is to get rid
of setup.cfg and leave only setup.py and pyproject.toml, in the hope that this simplifies
the debugging of various build dependency issues later.

Duncan Macleod authored May 28, 2021 and Sean Leavey committed Jun 04, 2021

inside a `if __name__ == "__main__":` clause

Duncan Macleod authored May 28, 2021 and Sean Leavey committed Jun 04, 2021

the build_ext command (now) natively supports parallel builds, so we just need to overload the default to use it, rather than use a homebrewed solution.
this can be configured/disabled using the `CPU_COUNT` environment variable, e.g. `CPU_COUNT=0 python3 build_ext`

- Sphinx now 4.*
- Fixed finesse_sphinx to work with Sphinx 4
- Updated sphinxcontrib-bibtex to 2.2.* to remove need for version pinning < 2.*
- Most other packages pinned to their latest PyPI versions
- Requirements files now don't pin versions for non-production tools (docs, build, etc.)

Store Cython extension sources as relative paths in setup.py

See merge request !78

1;95;0cMerge branch 'develop' of git.ligo.org:finesse/finesse3 into develop

so that the SOURCES.txt file generated by setuptools contains only relative paths - absolute paths can confuse windows

Don't open already open files in the test suite

Closes #361

See merge request !80

that doesn't work on windows

that doesn't work on windows

As discussed in the telecon

This dependency made it necessary for end users to install Cython even when using pre-built binary wheels.

The conda package will just be a wrapper around the PyPI wheels, and can't easily be tested
for each commit (yet) so removing for now.
Relevant chat: https://chat.ligo.org/ligo/pl/5t1uhntodbdapripqkntr6n6do