WantTensors=True, want_zstar=True causes error "binary_search (e.g for optical depth) did not converge"

[joezuntz]

Hi again,

Sorry for the issue flurry!

Running with both WantTensors=True and want_zstar=True at the same time seems to cause a binary_search error in calc_transfers. Not sure if this is expected for some reason.

Here's an example:

import camb
import traceback

for tensors in [False, True]:
    for zstar in [False, True]:

        cp = camb.set_params(ns=1, H0=67, ombh2=0.022, omch2=0.1,
                             lmax=2600, kmax=10.0, dark_energy_model='DarkEnergyPPF', 
                             WantTensors=tensors,
                             want_zstar=zstar)
        try:
            r = camb.get_results(cp)
            print("WantTensors={} and want_zstar={} succeeded".format(tensors, zstar))
        except camb.CAMBError:
            print("WantTensors={} and want_zstar={} failed with this error:".format(tensors, zstar))
            traceback.print_exc()

prints:

WantTensors=False and want_zstar=False succeeded
WantTensors=False and want_zstar=True succeeded
WantTensors=True and want_zstar=False succeeded
WantTensors=True and want_zstar=True failed with this error:
Traceback (most recent call last):
  File "tensors.py", line 12, in <module>
    r = camb.get_results(cp)
  File "/usr/local/lib/python3.7/site-packages/camb/camb.py", line 40, in get_results
    res.calc_power_spectra(params)
  File "/usr/local/lib/python3.7/site-packages/camb/results.py", line 307, in calc_power_spectra
    self.calc_transfers(params, only_transfers=False)
  File "/usr/local/lib/python3.7/site-packages/camb/results.py", line 290, in calc_transfers
    config.check_global_error('calc_transfer')
  File "/usr/local/lib/python3.7/site-packages/camb/_config.py", line 49, in check_global_error
    raise CAMBError(reference + '%s' % err)
camb.baseconfig.CAMBError: Error in Fortran called from calc_transfer:
binary_search (e.g for optical depth) did not converge

[cmbant]

Try 8b66482

[joezuntz]

Perfect fix - thanks v much!