Issues

• A failed NFsim load leaves the host process in the XML's directory: create_system() changes the cwd and an exception from initializeFromXML skips the restore. The stream swap and global RNG reseed also need to be explicit and contained

  bug
  lanl/bngsim#1314 · Filter by author wshlavacek opened on Oct 27, 2026

• Typed exceptions for "no sparse Jacobian for KLU" and "codegen declined this model"

  enhancement
  lanl/bngsim#1313 · Filter by author wshlavacek opened on Oct 27, 2026

• differential: NFsim (nf_reject) against RuleMonkey (nf_exact) in distribution

  enhancementgatedtesting
  lanl/bngsim#1312 · Filter by author wshlavacek opened on Oct 27, 2026

• differential: steady_state() against a long run to equilibrium

  enhancementgatedtesting
  lanl/bngsim#1311 · Filter by author wshlavacek opened on Oct 27, 2026

• differential: SSA with compiled propensities against interpreted ones, seed for seed

  enhancementgatedtesting
  lanl/bngsim#1310 · Filter by author wshlavacek opened on Oct 27, 2026

• differential: attribute a disagreement with a third arm, and flag answers that flip with the grid or tolerance

  enhancementgatedtesting
  lanl/bngsim#1309 · Filter by author wshlavacek opened on Oct 27, 2026

• differential: the analytic Jacobian's values against differences of the right-hand side, not only its pattern

  enhancementgatedtesting
  lanl/bngsim#1308 · Filter by author wshlavacek opened on Oct 27, 2026

• differential: state-leak checks: a restarted run, a reused Model or Simulator, and a batch against fresh serial runs

  enhancementgatedtesting
  lanl/bngsim#1307 · Filter by author wshlavacek opened on Oct 27, 2026

• differential: generate models (random networks and rate laws) and shrink each finding to a minimal reproducer

  enhancementgatedtesting
  lanl/bngsim#1306 · Filter by author wshlavacek opened on Oct 27, 2026

• differential: metamorphic transforms of the model, the check a shared loader cannot fail together

  enhancementgatedtesting
  lanl/bngsim#1305 · Filter by author wshlavacek opened on Oct 27, 2026

• BIOMD0000000338 and 339: KLU and the dense solver take different branches past a step-size collapse (Fg 311 on KLU, 0.73 on dense and libRoadRunner)

  bugcorrectness concern
  lanl/bngsim#1304 · Filter by author wshlavacek opened on Oct 27, 2026
  1

• sens: after a sensitivity-RHS refusal, the output loop reads the integrator outside its step (one read on a solver at t = 2.6e-312)

  lanl/bngsim#1301 · Filter by author Montana opened on Oct 27, 2026

• A sensitivity run's state departs from a converged plain run by 0.9% on MODEL2001200002 (no events), and does not converge onto it at 100x tighter tolerances

  lanl/bngsim#1298 · Filter by author Montana opened on Oct 26, 2026
  1

• BIOMD0000000810 asked for 8x the output points returns 5.3e7 of its trajectory's peak off the converged run, with no error

  lanl/bngsim#1297 · Filter by author Montana opened on Oct 26, 2026
  2

• jacobian="fd" alone is refused, telling the caller to drop force_sparse_linear_solver, which they never set: the automatic routing picked KLU, and the pattern has nothing to color

  lanl/bngsim#1296 · Filter by author Montana opened on Oct 26, 2026

• deps: move the uv.lock pins to antimony 3.2.1 and libroadrunner 2.11.0, after evaluating what each upgrade changes

  enhancementtesting
  lanl/bngsim#1288 · Filter by author wshlavacek opened on Oct 26, 2026

• A smooth narrow pulse (a Gaussian or sech² in a clock) is stepped over: Y = 0 for 2.212 at every tolerance, with no warning

  bugcorrectness concern
  lanl/bngsim#1286 · Filter by author wshlavacek opened on Oct 25, 2026
  1

• The clock-rate column of a narrow kinked pulse stops converging at 2e-5 to 9e-5 from rtol 1e-10 (w <= 0.005), where central differences of plain runs reach 1e-7

  bugfamily: sensitivity terms
  lanl/bngsim#1285 · Filter by author wshlavacek opened on Oct 25, 2026

• A rate-law equality the state sits on is judged at the run's start: an event that lands on it mid-run is still silent, and a stretch that outlasts the run is refused

  bugcorrectness concernfamily: sensitivity terms
  lanl/bngsim#1279 · Filter by author wshlavacek opened on Oct 25, 2026

• #1133's read-ahead still costs 1.15-1.34x at 20001 rows after #1200 (1.08-1.21x with sensitivities): read ahead only where a trigger can cross

  family: ODE eventsperformance
  lanl/bngsim#1278 · Filter by author wshlavacek opened on Oct 25, 2026

• A fixed-time event and a state trigger a column moves, on one instant, fire as two batches: S jumps 3.75 to 4.25 across k = 1 and the run returns dS/dk = 1.25 with no refusal

  bugcorrectness concernfamily: ODE eventsfamily: sensitivity terms
  lanl/bngsim#1277 · Filter by author wshlavacek opened on Oct 25, 2026

• A sensitivity run stopped by another root within 1e-12 before a sample still records that row before the event (#1182's read-ahead needs t_now == t_out)

  bugfamily: ODE eventsfamily: sensitivity termslow priority
  lanl/bngsim#1276 · Filter by author wshlavacek opened on Oct 25, 2026

• .net writer: an assignment-rule species is written as a $-fixed species beside a same-named observable or function; bngsim reads the rule, run_network holds the initial value with no message

  bugcorrectness concern
  lanl/bngsim#1274 · Filter by author Montana opened on Oct 25, 2026

• amici_sens: judge each sensitivity parameter on its own scale (after #1167)

  bugcorrectness concerntesting
  lanl/bngsim#1261 · Filter by author wshlavacek opened on Oct 24, 2026

• The analytic Jacobian of a size read through a diamond of rules takes exponential time to set up: 3.1 s at 12 rules, 35.8 s at 16, 53.5 s at 18, where jacobian="fd" builds in under 10 ms

  family: variable volumesperformance
  lanl/bngsim#1255 · Filter by author Montana opened on Oct 24, 2026