1.6.4

• Use pre-reionization x_e for baryon sound speed, avoiding unphysical result during reionization (fixes #171); still not correct as no realistic treatment of baryon pressure after reionization
• Added Linux arm wheels (e.g. for containers on Macs)
• Minor doc and version consistency updates

gfortran compiler bug giving memory leaks now fixed in gfortran 15.2, but CAMB continues to compile safely on older versions.

1.6.1

• Workaround for gfortran compiler memory leak bug affecting runs with tensors, thanks @eatdust for report
  (info and fixed gfortran tracking at #176)
• Updated for min Python 3.10 and latest setuptools
• Added formatting ruff, pre-commit hook and vscode formatting options

1.6.0

• doc clarifications
• Improved AI help assistant with full python doc knowledge
• Parameter to set max reionization redshift (thanks @mraveri , #166)
• Minor compatibility/parameter warning tweaks
• Initialize share_delta_neff back to False when used from, e.g. in case starting from copy of .ini file

1.5.7

• Change PPF switch parameter to give closer to fluid model in extreme modes; changes results, but very similar near LCDM (no unit tests fail)
• Minor doc/compatibility updates

1.5.5

• Added pre-compiled binary wheels for pip installation
• Doc updates including new section on code modification
• Minor fixes/tidies

1.5.3

• CAMBparams can now be saved and restored using repr/eval or pickle in standard cases
• Allow DE fluid w(a) table models with w<-1 as long as do not cross -1
• Removed camb.py
• Misc tidies and compatibility fixes

1.5.1

• Refactor reionization models allowing setting the model via python; added sample ExpReionization alternative to default tanh. Examples added to example notebook.
• Removed dependency on distutils ready for Python 12

1.5.0

• Fixed bug calculating Cl from time sources that affected iterated use in Cobaya
• Increase max transfer redshifts to 256
• Added min_l_logl_sampling parameter to set_accuracy (to set lowest multipole at which sparse log-L sampling is used)
• Added sample transfer function animations to the end of the example notebook

1.4.0

• Updated numerical constants, use latest definitions for Megaparsec, and calculate times in Julian giga-years
• Updated default nnu to 3.044 (from 3.046)
• Updated default BBN model (for helium abundance) to PRIMAT2021
• Support general bias(k,z) for source window functions
• Fix for HMcode-2020 feedback at low k
• Fix get_matter_power_spectrum for cross-spectra
• Allow read_ini to load directly from a URL
• Minor code tidies

These changes do change numerical results, but only at a level that is irrelevant for near-future observations. But some of the higher-precision unit tests also had to be updated.

1.3.6

• add optional nonlinear parameter to set_for_lmax
• add 21cm transfer variable names to python; add example notebook example
• recfast compatibility with very different T_0
• Minor code and doc fixes/warnings