Understand the failure.
Verify the repair.
Small, reproducible software failures. The broken implementation, the fix that didn’t work, and the one that passed—preserved together.
Explore the cases ↓How results are verified ↗WHAT THE ARCHIVE CONTAINS
100840 executable cases. 20168 are open.
Every case records the implementation that fails, the fix that did not work, and the repair that passed its checks—with recorded outputs and source hashes. This release adds 100840 cases across 20168 failure mechanisms and 254 domains.
The open tier gives you the failure and the unsuccessful fix for one case in every mechanism. The remaining 80672 cases, 5 variants per mechanism, are member-only: the verified repair, its recorded checks, and the full fixture suite are held in the member archive. Read the methodology ↗
A RECORD OF WHAT WENT WRONG
Browse the archive / 100840
RK4 second stage samples the forcing at the start of the step · case 01
Driven oscillators lag the drive phase and resonance peaks shift.
RK4 second stage samples the forcing at the start of the step · case 02
Driven oscillators lag the drive phase and resonance peaks shift.
RK4 second stage samples the forcing at the start of the step · case 03
Driven oscillators lag the drive phase and resonance peaks shift.
RK4 second stage samples the forcing at the start of the step · case 04
Driven oscillators lag the drive phase and resonance peaks shift.
RK4 second stage samples the forcing at the start of the step · case 05
Driven oscillators lag the drive phase and resonance peaks shift.
RK4 third stage reuses the first-stage slope · case 01
The method degrades to second order and energy error grows with dt^2.
RK4 third stage reuses the first-stage slope · case 02
The method degrades to second order and energy error grows with dt^2.
RK4 third stage reuses the first-stage slope · case 03
The method degrades to second order and energy error grows with dt^2.
RK4 third stage reuses the first-stage slope · case 04
The method degrades to second order and energy error grows with dt^2.
RK4 third stage reuses the first-stage slope · case 05
The method degrades to second order and energy error grows with dt^2.
RK4 fourth stage advances only half a step · case 01
End-of-step slope is evaluated at the midpoint and trajectories drift.
RK4 fourth stage advances only half a step · case 02
End-of-step slope is evaluated at the midpoint and trajectories drift.
RK4 fourth stage advances only half a step · case 03
End-of-step slope is evaluated at the midpoint and trajectories drift.
RK4 fourth stage advances only half a step · case 04
End-of-step slope is evaluated at the midpoint and trajectories drift.
RK4 fourth stage advances only half a step · case 05
End-of-step slope is evaluated at the midpoint and trajectories drift.
RK4 position update weights all stages equally · case 01
The combined step is only first order accurate.
RK4 position update weights all stages equally · case 02
The combined step is only first order accurate.
RK4 position update weights all stages equally · case 03
The combined step is only first order accurate.
RK4 position update weights all stages equally · case 04
The combined step is only first order accurate.
RK4 position update weights all stages equally · case 05
The combined step is only first order accurate.
RK4 loop never advances the simulation clock · case 01
The drive force is frozen at cos(0) and forced response is wrong.
RK4 loop never advances the simulation clock · case 02
The drive force is frozen at cos(0) and forced response is wrong.
RK4 loop never advances the simulation clock · case 03
The drive force is frozen at cos(0) and forced response is wrong.
RK4 loop never advances the simulation clock · case 04
The drive force is frozen at cos(0) and forced response is wrong.
RK4 loop never advances the simulation clock · case 05
The drive force is frozen at cos(0) and forced response is wrong.
Damping term in the RK4 derivative adds energy · case 01
A damped, driven system grows without bound.
Damping term in the RK4 derivative adds energy · case 02
A damped, driven system grows without bound.
Damping term in the RK4 derivative adds energy · case 03
A damped, driven system grows without bound.
Damping term in the RK4 derivative adds energy · case 04
A damped, driven system grows without bound.
Damping term in the RK4 derivative adds energy · case 05
A damped, driven system grows without bound.
Drive frequency is treated as hertz in an angular-frequency contract · case 01
The forcing oscillates 2*pi times too fast.
Drive frequency is treated as hertz in an angular-frequency contract · case 02
The forcing oscillates 2*pi times too fast.
Drive frequency is treated as hertz in an angular-frequency contract · case 03
The forcing oscillates 2*pi times too fast.
Drive frequency is treated as hertz in an angular-frequency contract · case 04
The forcing oscillates 2*pi times too fast.
Drive frequency is treated as hertz in an angular-frequency contract · case 05
The forcing oscillates 2*pi times too fast.
Step controller omits the error-order exponent on growth · case 01
Accepted steps grow erratically and trigger rejection cascades.
Step controller omits the error-order exponent on growth · case 02
Accepted steps grow erratically and trigger rejection cascades.
Step controller omits the error-order exponent on growth · case 03
Accepted steps grow erratically and trigger rejection cascades.
Step controller omits the error-order exponent on growth · case 04
Accepted steps grow erratically and trigger rejection cascades.
Step controller omits the error-order exponent on growth · case 05
Accepted steps grow erratically and trigger rejection cascades.
Adaptive integrator propagates the low-order Euler solution · case 01
The error estimate refers to Heun but the state follows Euler, so accuracy is worse than tol suggests.
Adaptive integrator propagates the low-order Euler solution · case 02
The error estimate refers to Heun but the state follows Euler, so accuracy is worse than tol suggests.
Adaptive integrator propagates the low-order Euler solution · case 03
The error estimate refers to Heun but the state follows Euler, so accuracy is worse than tol suggests.
Adaptive integrator propagates the low-order Euler solution · case 04
The error estimate refers to Heun but the state follows Euler, so accuracy is worse than tol suggests.
Adaptive integrator propagates the low-order Euler solution · case 05
The error estimate refers to Heun but the state follows Euler, so accuracy is worse than tol suggests.
Rejected steps still advance simulation time · case 01
The integrator reaches t_end early with the state lagging behind.
Rejected steps still advance simulation time · case 02
The integrator reaches t_end early with the state lagging behind.
Rejected steps still advance simulation time · case 03
The integrator reaches t_end early with the state lagging behind.
Rejected steps still advance simulation time · case 04
The integrator reaches t_end early with the state lagging behind.
Rejected steps still advance simulation time · case 05
The integrator reaches t_end early with the state lagging behind.
Adaptive steps overshoot the requested end time · case 01
The solution is reported for a time later than t_end.
Adaptive steps overshoot the requested end time · case 02
The solution is reported for a time later than t_end.
Adaptive steps overshoot the requested end time · case 03
The solution is reported for a time later than t_end.
Adaptive steps overshoot the requested end time · case 04
The solution is reported for a time later than t_end.
Adaptive steps overshoot the requested end time · case 05
The solution is reported for a time later than t_end.
Mixed error norm scales by the initial value instead of the current state · case 01
After the solution decays or grows the tolerance is effectively too tight or too loose.
Mixed error norm scales by the initial value instead of the current state · case 02
After the solution decays or grows the tolerance is effectively too tight or too loose.
Mixed error norm scales by the initial value instead of the current state · case 03
After the solution decays or grows the tolerance is effectively too tight or too loose.
Mixed error norm scales by the initial value instead of the current state · case 04
After the solution decays or grows the tolerance is effectively too tight or too loose.
Mixed error norm scales by the initial value instead of the current state · case 05
After the solution decays or grows the tolerance is effectively too tight or too loose.
Exact steps never grow the step size · case 01
A constant-slope phase keeps the tiny initial step forever and wastes work.
Exact steps never grow the step size · case 02
A constant-slope phase keeps the tiny initial step forever and wastes work.
Exact steps never grow the step size · case 03
A constant-slope phase keeps the tiny initial step forever and wastes work.
Exact steps never grow the step size · case 04
A constant-slope phase keeps the tiny initial step forever and wastes work.
Exact steps never grow the step size · case 05
A constant-slope phase keeps the tiny initial step forever and wastes work.
Rejected steps are always halved · case 01
Badly rejected steps retry too large and cause repeated rejections.
Rejected steps are always halved · case 02
Badly rejected steps retry too large and cause repeated rejections.
Rejected steps are always halved · case 03
Badly rejected steps retry too large and cause repeated rejections.
Rejected steps are always halved · case 04
Badly rejected steps retry too large and cause repeated rejections.
Rejected steps are always halved · case 05
Badly rejected steps retry too large and cause repeated rejections.
Leapfrog starts without offsetting velocity by half a step · case 01
The orbit phase is shifted and energy oscillation is biased from the first step.
Leapfrog starts without offsetting velocity by half a step · case 02
The orbit phase is shifted and energy oscillation is biased from the first step.
Leapfrog starts without offsetting velocity by half a step · case 03
The orbit phase is shifted and energy oscillation is biased from the first step.
Leapfrog starts without offsetting velocity by half a step · case 04
The orbit phase is shifted and energy oscillation is biased from the first step.
Leapfrog starts without offsetting velocity by half a step · case 05
The orbit phase is shifted and energy oscillation is biased from the first step.
Leapfrog reports the half-step velocity as the current velocity · case 01
Velocity output is offset by half a step, corrupting energy and contact logic.
Leapfrog reports the half-step velocity as the current velocity · case 02
Velocity output is offset by half a step, corrupting energy and contact logic.
Leapfrog reports the half-step velocity as the current velocity · case 03
Velocity output is offset by half a step, corrupting energy and contact logic.
Leapfrog reports the half-step velocity as the current velocity · case 04
Velocity output is offset by half a step, corrupting energy and contact logic.
Leapfrog reports the half-step velocity as the current velocity · case 05
Velocity output is offset by half a step, corrupting energy and contact logic.
Leapfrog kicks with the old position before drifting · case 01
The stored velocity lags a full step behind and the scheme loses time symmetry.
Leapfrog kicks with the old position before drifting · case 02
The stored velocity lags a full step behind and the scheme loses time symmetry.
Leapfrog kicks with the old position before drifting · case 03
The stored velocity lags a full step behind and the scheme loses time symmetry.
Leapfrog kicks with the old position before drifting · case 04
The stored velocity lags a full step behind and the scheme loses time symmetry.
Leapfrog kicks with the old position before drifting · case 05
The stored velocity lags a full step behind and the scheme loses time symmetry.
Constant field acceleration is added with the wrong sign · case 01
Bodies settle on the wrong side of the spring rest point.
Constant field acceleration is added with the wrong sign · case 02
Bodies settle on the wrong side of the spring rest point.
Constant field acceleration is added with the wrong sign · case 03
Bodies settle on the wrong side of the spring rest point.
Constant field acceleration is added with the wrong sign · case 04
Bodies settle on the wrong side of the spring rest point.
Constant field acceleration is added with the wrong sign · case 05
Bodies settle on the wrong side of the spring rest point.
Leapfrog drift adds a Taylor acceleration term on top of the staggered velocity · case 01
Acceleration is counted twice per step and trajectories overshoot.
Leapfrog drift adds a Taylor acceleration term on top of the staggered velocity · case 02
Acceleration is counted twice per step and trajectories overshoot.
Leapfrog drift adds a Taylor acceleration term on top of the staggered velocity · case 03
Acceleration is counted twice per step and trajectories overshoot.
Leapfrog drift adds a Taylor acceleration term on top of the staggered velocity · case 04
Acceleration is counted twice per step and trajectories overshoot.
Leapfrog drift adds a Taylor acceleration term on top of the staggered velocity · case 05
Acceleration is counted twice per step and trajectories overshoot.
Backward Euler denominator omits the implicit stiffness term · case 01
Stiff springs at large dt overshoot and gain energy, defeating the implicit scheme.
Backward Euler denominator omits the implicit stiffness term · case 02
Stiff springs at large dt overshoot and gain energy, defeating the implicit scheme.
Backward Euler denominator omits the implicit stiffness term · case 03
Stiff springs at large dt overshoot and gain energy, defeating the implicit scheme.
Backward Euler denominator omits the implicit stiffness term · case 04
Stiff springs at large dt overshoot and gain energy, defeating the implicit scheme.
Backward Euler denominator omits the implicit stiffness term · case 05
Stiff springs at large dt overshoot and gain energy, defeating the implicit scheme.
INSPECTABLE BY DESIGN
Every result has a runnable source.
Runnable implementations with recorded outputs, source hashes, and explicit contracts. Related variants share a failure mechanism and belong together in evaluation splits.
Read the methodology ↗