flatsurf / e-antic

Embedded algebraic number fields
https://flatsurf.github.io/e-antic/libeantic/
GNU Lesser General Public License v3.0
12 stars 11 forks source link

Test also with latest versions of FLINT in CI #249

Closed saraedum closed 1 year ago

saraedum commented 2 years ago

Additionally removed references to the flatsurf channel when installing dependencies. We do not want to install dependencies from our nightly channel anymore. Everything we need is now on conda-forge. There are some very old builds on our nightly channels with broken pins that sometimes get pulled in incorrectly.

Dependencies
codecov[bot] commented 2 years ago

Codecov Report

Merging #249 (1e81e3b) into master (0a17b73) will not change coverage. The diff coverage is n/a.

@@           Coverage Diff           @@
##           master     #249   +/-   ##
=======================================
  Coverage   92.04%   92.04%           
=======================================
  Files         105      105           
  Lines        2075     2075           
=======================================
  Hits         1910     1910           
  Misses        165      165           

:mega: We’re building smart automated test selection to slash your CI/CD build times. Learn more

saraedum commented 2 years ago

The "undefined behavior" that blocked some of the checks was only fixed in 2.9.0. So we need that to test things.

Packaged in https://github.com/conda-forge/libflint-feedstock/pull/27.

saraedum commented 2 years ago

The pyeantic test fails because a missing dependency on ipython_genutils. ipython_genutils used to be a dependency of traitlets but is not anymore. So ipykernel did not explicitly depend on ipython_genutils even though it used it because it got it from traitlets. When traitlets dropped their dependency on ipython_genutils, the dependency was explicitly added at some point in ipykernel but that change was not made to the pinning of ipykernel. So, an old ipykernel now pulls in a new traitlets which does not pull in ipython_genutils which that old ipykernel needs.

This should be fixed with a repodata patch to conda-forge.

The dependency was made explicit in https://github.com/ipython/ipykernel/pull/778 and then fully removed at some point.