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
Relative period error has the reversed sign · case 01
A lagging integrator is reported as leading.
Relative period error has the reversed sign · case 02
A lagging integrator is reported as leading.
Relative period error has the reversed sign · case 03
A lagging integrator is reported as leading.
Relative period error has the reversed sign · case 04
A lagging integrator is reported as leading.
Relative period error has the reversed sign · case 05
A lagging integrator is reported as leading.
Crossing that lands exactly on a zero sample is missed · case 01
Sampled trajectories with exact zeros report missing periods.
Crossing that lands exactly on a zero sample is missed · case 02
Sampled trajectories with exact zeros report missing periods.
Crossing that lands exactly on a zero sample is missed · case 03
Sampled trajectories with exact zeros report missing periods.
Crossing that lands exactly on a zero sample is missed · case 04
Sampled trajectories with exact zeros report missing periods.
Crossing that lands exactly on a zero sample is missed · case 05
Sampled trajectories with exact zeros report missing periods.
Energy drift is measured step-to-step instead of against the initial energy · case 01
Slow monotone drift never trips the monitor.
Energy drift is measured step-to-step instead of against the initial energy · case 02
Slow monotone drift never trips the monitor.
Energy drift is measured step-to-step instead of against the initial energy · case 03
Slow monotone drift never trips the monitor.
Energy drift is measured step-to-step instead of against the initial energy · case 04
Slow monotone drift never trips the monitor.
Energy drift is measured step-to-step instead of against the initial energy · case 05
Slow monotone drift never trips the monitor.
Energy loss is not treated as drift · case 01
Numerically dissipative integrators pass the monitor.
Energy loss is not treated as drift · case 02
Numerically dissipative integrators pass the monitor.
Energy loss is not treated as drift · case 03
Numerically dissipative integrators pass the monitor.
Energy loss is not treated as drift · case 04
Numerically dissipative integrators pass the monitor.
Energy loss is not treated as drift · case 05
Numerically dissipative integrators pass the monitor.
Zero initial energy is scaled by a tiny epsilon · case 01
A system starting at rest reports astronomically large drift for any motion.
Zero initial energy is scaled by a tiny epsilon · case 02
A system starting at rest reports astronomically large drift for any motion.
Zero initial energy is scaled by a tiny epsilon · case 03
A system starting at rest reports astronomically large drift for any motion.
Zero initial energy is scaled by a tiny epsilon · case 04
A system starting at rest reports astronomically large drift for any motion.
Zero initial energy is scaled by a tiny epsilon · case 05
A system starting at rest reports astronomically large drift for any motion.
Monitor returns drift at the first exceedance before scanning for blowup · case 01
An exploding simulation is reported as mild drift.
Monitor returns drift at the first exceedance before scanning for blowup · case 02
An exploding simulation is reported as mild drift.
Monitor returns drift at the first exceedance before scanning for blowup · case 03
An exploding simulation is reported as mild drift.
Monitor returns drift at the first exceedance before scanning for blowup · case 04
An exploding simulation is reported as mild drift.
Monitor returns drift at the first exceedance before scanning for blowup · case 05
An exploding simulation is reported as mild drift.
Blowup factor is applied to the deviation rather than the energy · case 01
Explosions are detected one multiple of E0 too late.
Blowup factor is applied to the deviation rather than the energy · case 02
Explosions are detected one multiple of E0 too late.
Blowup factor is applied to the deviation rather than the energy · case 03
Explosions are detected one multiple of E0 too late.
Blowup factor is applied to the deviation rather than the energy · case 04
Explosions are detected one multiple of E0 too late.
Blowup factor is applied to the deviation rather than the energy · case 05
Explosions are detected one multiple of E0 too late.
Drift index reports the last exceedance · case 01
Diagnostics point at the end of the run instead of where drift began.
Drift index reports the last exceedance · case 02
Diagnostics point at the end of the run instead of where drift began.
Drift index reports the last exceedance · case 03
Diagnostics point at the end of the run instead of where drift began.
Drift index reports the last exceedance · case 04
Diagnostics point at the end of the run instead of where drift began.
Drift index reports the last exceedance · case 05
Diagnostics point at the end of the run instead of where drift began.
Sweep processes end events before begin events at equal x · case 01
Boxes whose x extents only touch never become a pair.
Sweep processes end events before begin events at equal x · case 02
Boxes whose x extents only touch never become a pair.
Sweep processes end events before begin events at equal x · case 03
Boxes whose x extents only touch never become a pair.
Sweep processes end events before begin events at equal x · case 04
Boxes whose x extents only touch never become a pair.
Sweep processes end events before begin events at equal x · case 05
Boxes whose x extents only touch never become a pair.
Secondary axis test is exclusive at the boundary · case 01
Stacked boxes resting exactly on each other lose their pair.
Secondary axis test is exclusive at the boundary · case 02
Stacked boxes resting exactly on each other lose their pair.
Secondary axis test is exclusive at the boundary · case 03
Stacked boxes resting exactly on each other lose their pair.
Secondary axis test is exclusive at the boundary · case 04
Stacked boxes resting exactly on each other lose their pair.
Secondary axis test is exclusive at the boundary · case 05
Stacked boxes resting exactly on each other lose their pair.
Sweep reports pairs in active-list order · case 01
The pair cache sees (b,a) one frame and (a,b) the next, restarting contacts.
Sweep reports pairs in active-list order · case 02
The pair cache sees (b,a) one frame and (a,b) the next, restarting contacts.
Sweep reports pairs in active-list order · case 03
The pair cache sees (b,a) one frame and (a,b) the next, restarting contacts.
Sweep reports pairs in active-list order · case 04
The pair cache sees (b,a) one frame and (a,b) the next, restarting contacts.
Sweep reports pairs in active-list order · case 05
The pair cache sees (b,a) one frame and (a,b) the next, restarting contacts.
Ended boxes stay in the active list · case 01
Boxes far apart on x are reported as candidates.
Ended boxes stay in the active list · case 02
Boxes far apart on x are reported as candidates.
Ended boxes stay in the active list · case 03
Boxes far apart on x are reported as candidates.
Ended boxes stay in the active list · case 04
Boxes far apart on x are reported as candidates.
Ended boxes stay in the active list · case 05
Boxes far apart on x are reported as candidates.
Secondary test reads max x as max y · case 01
Tall thin boxes miss pairs and wide flat boxes gain false ones.
Secondary test reads max x as max y · case 02
Tall thin boxes miss pairs and wide flat boxes gain false ones.
Secondary test reads max x as max y · case 03
Tall thin boxes miss pairs and wide flat boxes gain false ones.
Secondary test reads max x as max y · case 04
Tall thin boxes miss pairs and wide flat boxes gain false ones.
Secondary test reads max x as max y · case 05
Tall thin boxes miss pairs and wide flat boxes gain false ones.
Grid cell index truncates toward zero for negative coordinates · case 01
Boxes straddling the origin share cell 0 with boxes one cell away.
Grid cell index truncates toward zero for negative coordinates · case 02
Boxes straddling the origin share cell 0 with boxes one cell away.
Grid cell index truncates toward zero for negative coordinates · case 03
Boxes straddling the origin share cell 0 with boxes one cell away.
Grid cell index truncates toward zero for negative coordinates · case 04
Boxes straddling the origin share cell 0 with boxes one cell away.
Grid cell index truncates toward zero for negative coordinates · case 05
Boxes straddling the origin share cell 0 with boxes one cell away.
Grid insertion excludes the cell containing the maximum corner · case 01
Boxes are missing from their last row and column of cells and contacts there are lost.
Grid insertion excludes the cell containing the maximum corner · case 02
Boxes are missing from their last row and column of cells and contacts there are lost.
Grid insertion excludes the cell containing the maximum corner · case 03
Boxes are missing from their last row and column of cells and contacts there are lost.
Grid insertion excludes the cell containing the maximum corner · case 04
Boxes are missing from their last row and column of cells and contacts there are lost.
Grid insertion excludes the cell containing the maximum corner · case 05
Boxes are missing from their last row and column of cells and contacts there are lost.
Candidate pairs found in several cells are reported repeatedly · case 01
Narrowphase runs the same pair once per shared cell.
Candidate pairs found in several cells are reported repeatedly · case 02
Narrowphase runs the same pair once per shared cell.
Candidate pairs found in several cells are reported repeatedly · case 03
Narrowphase runs the same pair once per shared cell.
Candidate pairs found in several cells are reported repeatedly · case 04
Narrowphase runs the same pair once per shared cell.
Candidate pairs found in several cells are reported repeatedly · case 05
Narrowphase runs the same pair once per shared cell.
Grid row range uses the maximum x as maximum y · case 01
Tall boxes are inserted into too few rows, wide boxes into too many.
Grid row range uses the maximum x as maximum y · case 02
Tall boxes are inserted into too few rows, wide boxes into too many.
Grid row range uses the maximum x as maximum y · case 03
Tall boxes are inserted into too few rows, wide boxes into too many.
Grid row range uses the maximum x as maximum y · case 04
Tall boxes are inserted into too few rows, wide boxes into too many.
Grid row range uses the maximum x as maximum y · case 05
Tall boxes are inserted into too few rows, wide boxes into too many.
Largest bucket statistic reports the number of occupied cells · case 01
Grid tuning picks cell sizes from a meaningless load metric.
Largest bucket statistic reports the number of occupied cells · case 02
Grid tuning picks cell sizes from a meaningless load metric.
Largest bucket statistic reports the number of occupied cells · case 03
Grid tuning picks cell sizes from a meaningless load metric.
Largest bucket statistic reports the number of occupied cells · case 04
Grid tuning picks cell sizes from a meaningless load metric.
Largest bucket statistic reports the number of occupied cells · case 05
Grid tuning picks cell sizes from a meaningless load metric.
Fat box containment is strict · case 01
A proxy whose tight box touches its fat bound is reinserted every frame.
Fat box containment is strict · case 02
A proxy whose tight box touches its fat bound is reinserted every frame.
Fat box containment is strict · case 03
A proxy whose tight box touches its fat bound is reinserted every frame.
Fat box containment is strict · case 04
A proxy whose tight box touches its fat bound is reinserted every frame.
Fat box containment is strict · case 05
A proxy whose tight box touches its fat bound is reinserted every frame.
Leftward motion extends the right side of the fat box · case 01
Proxies moving left escape their fat box immediately and churn the tree.
Leftward motion extends the right side of the fat box · case 02
Proxies moving left escape their fat box immediately and churn the tree.
Leftward motion extends the right side of the fat box · case 03
Proxies moving left escape their fat box immediately and churn the tree.
Leftward motion extends the right side of the fat box · case 04
Proxies moving left escape their fat box immediately and churn the tree.
Leftward motion extends the right side of the fat box · case 05
Proxies moving left escape their fat box immediately and churn the tree.
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 ↗