[ColmTalbot]

BaseSmoothedMassDistribution gains ln_prob, ln_p_m1, ln_p_q and ln_smoothing, and an ln_primary_model class attribute, the log of primary_model taking the same arguments (the log of primary_model if unset); its four subclasses and InterpolatedPowerlaw set it. The terms for each sample are computed in log space, the smoothing window as a log-sigmoid, -logaddexp(0, exponent); the normalizations are computed as for call. InterpolatedNoBaseModelIdentical, and so the spline spin models, gains ln_prob, ln_p_x_identical and ln_p_x_unnormed: the spline perturbation minus the log of its normalization.

The linear path is refactored, with identical values:

• the mmin/mmax check moves to _check_mass_limits, and its error messages are now f-strings (they were printed unformatted);
• the cached mass ratio normalization moves to _norm_p_q_at, shared by p_q and ln_p_q, which also re-caches when the cached interpolant's shape differs from mass_1 (before, a shape that still broadcast silently used stale normalizations);
• the smoothing exponent moves to _smoothing_exponent.

In a float32 survey (10k samples x 10 draws per model), the points lost to underflow drop from 0.3-2.6% to almost none. The exception is primary masses within ~delta_m / 87 of mmin, where the linear mass ratio normalization itself underflows (7 of 500k points, all with ln p below -170). Close to mmin the float32 error grows like (ln p)^2, as the smoothing is evaluated at m - mmin; it is below 5e-3 where ln p > -84.

[codecov]

Codecov Report

❌ Patch coverage is 95.50562% with 4 lines in your changes missing coverage. Please review.
✅ Project coverage is 95.61%. Comparing base (5dfefe4) to head (c8f2e74).

Files with missing lines	Patch %	Lines
gwpopulation/models/mass.py	94.52%	4 Missing ⚠️

Additional details and impacted files

@@                Coverage Diff                 @@
##           ln-mass-models     #148      +/-   ##
==================================================
+ Coverage           95.50%   95.61%   +0.11%     
==================================================
  Files                  14       14              
  Lines                1156     1230      +74     
==================================================
+ Hits                 1104     1176      +72     
- Misses                 52       54       +2     

Flag	Coverage Δ
python3.10	95.61% <95.51%> (+0.11%)	⬆️
python3.11	95.61% <95.51%> (+0.11%)	⬆️
python3.12	95.61% <95.51%> (+0.11%)	⬆️
python3.13	95.61% <95.51%> (+0.11%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.