• Same as v1.2.0 but with new wheels for Python 3.10 and AArch64.

• Same as v1.1.0 but with wheels built and uploaded to PyPI. Please report if you
  have problems with wheels.

• Add segmentation masking to the photometry and kron/auto functions (#69).

• Add functions sep.set_sub_object_limit(limit) and sep.get_sub_object_limit()
  for modifying and retrieving the sub-object deblending limit. Previously this
  parameter was hard-coded to 1024. 1024 is now the default value. (#84)

• This and future versions are now Python 3 only. Python 2 is no longer
  supported.

• Modernize setup.py with pyproject.toml

• Fix makefile so that make install works on OS X for the C library.
  Python module and C code are unchanged.

• Fix bug when using masked filter and noise array where objects with member pixels at end of image (maximum y coordinate) were erroneously missed.

This release removes features deprecated in previous versions. All future v1.x.x versions will be backwards compatible. It also includes one bugfix:

• Fix bug in Background.rms() giving nonsensical results.

• New, more coherent C API. This change should be transparent to users
  of the Python module.
• Add variance uncertainty parameters errx2, erry2 and errxy to
  output of sep.extract().
• Add a minimum sigma to sep.winpos() to match Source Extractor
  behavior.
• Fix use of boolean masks in sep.kron_radius(). Formerly, using a
  boolean mask resulted in nonsense results.
• Fix segfault in Background.back() when box size is same as image size.
• Fix bug in creating long error messages on Python 3.

Adds support for OS X and Windows.