Fix CI, installs failing caused by numpy API change #148
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Without this option, pip will go fetch all the install_requires listed in setup.py for an isolated build. The problem is that the version of numpy it grabs could be different to one that is _already_ present in the conda env. Since numpy _is_ already present in the conda env, we expose it to pip using the --no-build-isolation flag. This way, our C extension is built against the same numpy as will be present in the conda env. This is particularly important given numpy 1.20 has recently shifted API version (see numpy/numpy/core/include/numpy/numpyconfig.h, numpy/numpy/core/code_generators/cversions.txt).
First, note that the requirements for the build environment (e.g. for extensions) is different to the requirements for running the package (i.e. install_requires). Build requirements are listed in pyproject.toml, install requirements in setup.py. We don't need --no-build-isolation if we specify the same numpy requirements in both. As such, I have reverted the previous commit. There is a consistent numpy C-API between numpy 1.17 (technically 1.16.1) and 1.20 (see previous commit). This commit consistently limits the numpy versions to that range.
… process" Cleanup after debug. This reverts commit a7ee1fe.
Cleanup before merge. This reverts commit 321fd05.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Pull Request Checklist
devbranch?Exceptions will be made for urgent bugfixes.
dev?If not, please rebase your PR onto the most recent
devtip.Explain to a new user by completing the sentence: 'This PR will: ...'
Description of the Change
qFit seems to compile and install fine.
On some systems though (python >= 3.7), we might get a RuntimeError on any attempt to run a qFit program:
RuntimeError: module compiled against API version 0xe but this version of numpy is 0xd.This is caused by different build and install requirements.
There is a consistent API version between numpy>=1.17,<1.20 (API 0xd).
This PR limits the numpy versions to this range in both the build (pyproject.toml) and install requirements (setup.py).
Release Notes