Since #208 our codebuilds on AWS have failed and a local build of the genet docker image fails. This is caused by pip/setuptools hanging during the package installation.
I have tried multiple approaches to debug / fix this and have not yet found a solution. Attempted:
Remove all genet dependencies.
Remove all pip args (--no-cache-dir, --compile) from Dockerfile command.
Use python -m build / pip wheel instead of pip install.
The issue occurs when "Getting build dependencies for sdist" (build) / "Getting requirements to build wheel" (pip install) but that is as verbose as the log is willing to be. I have not found this issue registered in any of the pypa github repositories (build, setuptools, pip) and there are no obvious forum discussions on the topic.
What's particularly weird about it is that we have no problems with the installation outside Docker. This step takes less than a second on my Mac.
I'll update on this as soon as I have news from my query on the pip GH issue tracker.
Solved by some helpful people over in https://github.com/pypa/pip/issues/12426. Turns out that installing from the root directory causes some library to become confused and it hurts itself in its confusion.
Since #208 our codebuilds on AWS have failed and a local build of the genet docker image fails. This is caused by pip/setuptools hanging during the package installation.
I have tried multiple approaches to debug / fix this and have not yet found a solution. Attempted:
--no-cache-dir
,--compile
) from Dockerfile command.python -m build
/pip wheel
instead ofpip install
.The issue occurs when "Getting build dependencies for sdist" (
build
) / "Getting requirements to build wheel" (pip install
) but that is as verbose as the log is willing to be. I have not found this issue registered in any of thepypa
github repositories (build
,setuptools
,pip
) and there are no obvious forum discussions on the topic.What's particularly weird about it is that we have no problems with the installation outside Docker. This step takes less than a second on my Mac.
I'll update on this as soon as I have news from my query on the pip GH issue tracker.