[tompollard]

calc_adc_params checks for infinite values before computing the ADC gain and baseline. For p_signal this uses np.isinf, but for e_p_signal it only compared each channel's minimum and maximum against math.inf. A channel containing -inf got past the check and failed later with a misleading error:

import numpy as np, wfdb
wfdb.Record(e_p_signal=[np.array([-np.inf, 1.0])], fmt=["16"]).calc_adc_params()
# ValueError: cannot convert float NaN to integer

It now uses math.isinf, so -inf raises "Signal contains inf. Cannot perform adc." like the p_signal branch.

Testing:

• New TestSignal.test_adc_params_inf covers +inf and -inf in both p_signal and e_p_signal. The -inf / e_p_signal case fails on main and passes with this change.
• uv run pytest -m "not network": 349 passed.
• ruff format --check and ruff check pass. Changelog entry added in changelog.d/.