lfungenus2 [5].lfungenus2 says the model must be minimal at $2$ and, when the conductor is even, the conductor's valuation at $2$ may be incorrect [5].default(realbitprecision, 320);
P = x^5 + x^4;
Q = x^3 + x + 1;
L = lfungenus2([P, Q]); \\ L-function of 169.a.169.1
r = lfunorderzero(L, 1);
lfun(L, 1, r) / r! \\ 0.090490390832429629113589757258...
lfuncheckfeq(L) \\ about -320The generator uses the $41$ odd-conductor LMFDB genus $2$ isogeny classes with conductor $N\leq1000$, grouped from $56$ curve rows. For each class it builds the PARI L-function [5] with lfungenus2([P,Q]), where PARI's $[P,Q]$ is the model $[f,h]$ in $y^2+h(x)y=f(x)$.
It computes lfun(L,1,r,bitprec)/r! at $60$ and $80$ decimal working digits using explicit bit precision, and stores the digits that agree. Before the draft was created, PARI's lfunorderzero agreed with the LMFDB analytic rank on all $41$ classes, lfuncheckfeq(L,None,330) returned between $-339$ and $-331$, and the computed values agreed with the LMFDB leading_coeff fields, including [4]. Even-conductor classes were left out because their local data at $2$ need separate checking before the same computation is used. The BSD quotient from the LMFDB period, regulator, Tamagawa product, analytic order of $\text{Sha}$, and torsion order agreed to the precision those source fields carry.