Deploying rubin-env 6.0.0

On Tuesday, 2023-04-18, we updated our build infrastructure to use version 6.0.0 of the rubin-env conda-forge metapackage. The d_2023_02_19 daily and w_2023_16 weekly will be the first tagged releases to incorporate this change.

This new environment adds the torchvision (RFC-909) package. It removes the mpich package that is no longer used within the Science Pipelines. Several packages that had been pinned to maximum versions due to incompatibilities have had those pins released: coverage, lmfit (DM-32406), pandas (DM-38549 and DM-38553), photutils (DM-38622), scikit-image (DM-38575 and DM-38582), and sqlalchemy (DM-30178, DM-31589, and DM-38678). Minimum pins for astropy (>=5.2), photutils (>=1.7), and scipy (>=1.10) have been established due to requirements for features. The Linux C, C++, and FORTRAN compiler versions were updated from 10 to 11, and llvm_openmp was updated from version 13 to 14 in order to follow the central conda-forge pinnings. numpy was updated to 1.23, which is the latest that we believe is compatible with numba with Python 3.10.

Iā€™d like to thank everyone who helped with the above tickets and especially @erykoff and @timj for helping get this out the door.

The rubin-env-developer metapackage had types-click removed, as it now conflicts with the click package.

Source and binary packages that built under rubin-env 5.1.0 will not generally function under rubin-env 6.0.0, as indicated by the semantic versioning. Despite the new photutils requirement, backward compatibility code was inserted so that weekly release w_2022_16 will build under rubin-env 5.1.0, but later weeklies may not.

1 Like