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
Ranked alpha clip coverage: rotation sign · case 01
The coverage dither rotation moves selected sample IDs in the opposite direction.
Ranked alpha clip coverage: rotation sign · case 02
The coverage dither rotation moves selected sample IDs in the opposite direction.
Ranked alpha clip coverage: rotation sign · case 03
The coverage dither rotation moves selected sample IDs in the opposite direction.
Ranked alpha clip coverage: rotation sign · case 04
The coverage dither rotation moves selected sample IDs in the opposite direction.
Ranked alpha clip coverage: rotation sign · case 05
The coverage dither rotation moves selected sample IDs in the opposite direction.
Ranked alpha clip coverage: rotation period · case 01
Rotation wraps around the covered-sample count rather than physical sample count.
Ranked alpha clip coverage: rotation period · case 02
Rotation wraps around the covered-sample count rather than physical sample count.
Ranked alpha clip coverage: rotation period · case 03
Rotation wraps around the covered-sample count rather than physical sample count.
Ranked alpha clip coverage: rotation period · case 04
Rotation wraps around the covered-sample count rather than physical sample count.
Ranked alpha clip coverage: rotation period · case 05
Rotation wraps around the covered-sample count rather than physical sample count.
Ranked alpha clip coverage: disabled alpha · case 01
Disabling alpha-to-coverage produces no coverage instead of passing all samples.
Ranked alpha clip coverage: disabled alpha · case 02
Disabling alpha-to-coverage produces no coverage instead of passing all samples.
Ranked alpha clip coverage: disabled alpha · case 03
Disabling alpha-to-coverage produces no coverage instead of passing all samples.
Ranked alpha clip coverage: disabled alpha · case 04
Disabling alpha-to-coverage produces no coverage instead of passing all samples.
Ranked alpha clip coverage: disabled alpha · case 05
Disabling alpha-to-coverage produces no coverage instead of passing all samples.
Ranked alpha clip coverage: shader empty · case 01
An explicit empty shader clip mask is confused with a missing mask.
Ranked alpha clip coverage: shader empty · case 02
An explicit empty shader clip mask is confused with a missing mask.
Ranked alpha clip coverage: shader empty · case 03
An explicit empty shader clip mask is confused with a missing mask.
Ranked alpha clip coverage: shader empty · case 04
An explicit empty shader clip mask is confused with a missing mask.
Ranked alpha clip coverage: shader empty · case 05
An explicit empty shader clip mask is confused with a missing mask.
Ranked alpha clip coverage: geometry combine · case 01
Geometric coverage is unioned with alpha coverage instead of constraining it.
Ranked alpha clip coverage: geometry combine · case 02
Geometric coverage is unioned with alpha coverage instead of constraining it.
Ranked alpha clip coverage: geometry combine · case 03
Geometric coverage is unioned with alpha coverage instead of constraining it.
Ranked alpha clip coverage: geometry combine · case 04
Geometric coverage is unioned with alpha coverage instead of constraining it.
Ranked alpha clip coverage: geometry combine · case 05
Geometric coverage is unioned with alpha coverage instead of constraining it.
Ranked alpha clip coverage: shader fixed combine · case 01
Fixed-function and shader masks are unioned so either can re-enable a disabled sample.
Ranked alpha clip coverage: shader fixed combine · case 02
Fixed-function and shader masks are unioned so either can re-enable a disabled sample.
Ranked alpha clip coverage: shader fixed combine · case 03
Fixed-function and shader masks are unioned so either can re-enable a disabled sample.
Ranked alpha clip coverage: shader fixed combine · case 04
Fixed-function and shader masks are unioned so either can re-enable a disabled sample.
Ranked alpha clip coverage: shader fixed combine · case 05
Fixed-function and shader masks are unioned so either can re-enable a disabled sample.
Cohen sutherland pixel window: right edge inclusive · case 01
A pixel center exactly on the last column is classified as outside, so clipping it back onto the same edge never clears the outcode and the segment is dropped.
Cohen sutherland pixel window: right edge inclusive · case 02
A pixel center exactly on the last column is classified as outside, so clipping it back onto the same edge never clears the outcode and the segment is dropped.
Cohen sutherland pixel window: right edge inclusive · case 03
A pixel center exactly on the last column is classified as outside, so clipping it back onto the same edge never clears the outcode and the segment is dropped.
Cohen sutherland pixel window: right edge inclusive · case 04
A pixel center exactly on the last column is classified as outside, so clipping it back onto the same edge never clears the outcode and the segment is dropped.
Cohen sutherland pixel window: right edge inclusive · case 05
A pixel center exactly on the last column is classified as outside, so clipping it back onto the same edge never clears the outcode and the segment is dropped.
Cohen sutherland pixel window: trivial reject any outside · case 01
Trivial rejection fires whenever both endpoints are outside, discarding segments that cross the whole window between two different outside regions.
Cohen sutherland pixel window: trivial reject any outside · case 02
Trivial rejection fires whenever both endpoints are outside, discarding segments that cross the whole window between two different outside regions.
Cohen sutherland pixel window: trivial reject any outside · case 03
Trivial rejection fires whenever both endpoints are outside, discarding segments that cross the whole window between two different outside regions.
Cohen sutherland pixel window: trivial reject any outside · case 04
Trivial rejection fires whenever both endpoints are outside, discarding segments that cross the whole window between two different outside regions.
Cohen sutherland pixel window: trivial reject any outside · case 05
Trivial rejection fires whenever both endpoints are outside, discarding segments that cross the whole window between two different outside regions.
Cohen sutherland pixel window: outside code merge · case 01
The edge to clip is chosen from the union of both outcodes, so an intersection computed for one endpoint is written into the other endpoint.
Cohen sutherland pixel window: outside code merge · case 02
The edge to clip is chosen from the union of both outcodes, so an intersection computed for one endpoint is written into the other endpoint.
Cohen sutherland pixel window: outside code merge · case 03
The edge to clip is chosen from the union of both outcodes, so an intersection computed for one endpoint is written into the other endpoint.
Cohen sutherland pixel window: outside code merge · case 04
The edge to clip is chosen from the union of both outcodes, so an intersection computed for one endpoint is written into the other endpoint.
Cohen sutherland pixel window: outside code merge · case 05
The edge to clip is chosen from the union of both outcodes, so an intersection computed for one endpoint is written into the other endpoint.
Cohen sutherland pixel window: top edge parameter origin · case 01
The top-edge intersection measures the parameter from the far endpoint while interpolating from the near endpoint.
Cohen sutherland pixel window: top edge parameter origin · case 02
The top-edge intersection measures the parameter from the far endpoint while interpolating from the near endpoint.
Cohen sutherland pixel window: top edge parameter origin · case 03
The top-edge intersection measures the parameter from the far endpoint while interpolating from the near endpoint.
Cohen sutherland pixel window: top edge parameter origin · case 04
The top-edge intersection measures the parameter from the far endpoint while interpolating from the near endpoint.
Cohen sutherland pixel window: top edge parameter origin · case 05
The top-edge intersection measures the parameter from the far endpoint while interpolating from the near endpoint.
Cohen sutherland pixel window: stale outcode after clip · case 01
After moving endpoint 0 onto a window edge its outcode is assumed to be zero, so a point pushed into a different outside region is treated as inside.
Cohen sutherland pixel window: stale outcode after clip · case 02
After moving endpoint 0 onto a window edge its outcode is assumed to be zero, so a point pushed into a different outside region is treated as inside.
Cohen sutherland pixel window: stale outcode after clip · case 03
After moving endpoint 0 onto a window edge its outcode is assumed to be zero, so a point pushed into a different outside region is treated as inside.
Cohen sutherland pixel window: stale outcode after clip · case 04
After moving endpoint 0 onto a window edge its outcode is assumed to be zero, so a point pushed into a different outside region is treated as inside.
Cohen sutherland pixel window: stale outcode after clip · case 05
After moving endpoint 0 onto a window edge its outcode is assumed to be zero, so a point pushed into a different outside region is treated as inside.
Cohen sutherland pixel window: accept rounding mode · case 01
Accepted fractional endpoints are truncated instead of rounded half-up to the nearest pixel center.
Cohen sutherland pixel window: accept rounding mode · case 02
Accepted fractional endpoints are truncated instead of rounded half-up to the nearest pixel center.
Cohen sutherland pixel window: accept rounding mode · case 03
Accepted fractional endpoints are truncated instead of rounded half-up to the nearest pixel center.
Cohen sutherland pixel window: accept rounding mode · case 04
Accepted fractional endpoints are truncated instead of rounded half-up to the nearest pixel center.
Cohen sutherland pixel window: accept rounding mode · case 05
Accepted fractional endpoints are truncated instead of rounded half-up to the nearest pixel center.
Cohen sutherland pixel window: window uses pixel count · case 01
The window maximum uses the pixel count rather than the last pixel center, letting clipped endpoints land one pixel beyond the raster.
Cohen sutherland pixel window: window uses pixel count · case 02
The window maximum uses the pixel count rather than the last pixel center, letting clipped endpoints land one pixel beyond the raster.
Cohen sutherland pixel window: window uses pixel count · case 03
The window maximum uses the pixel count rather than the last pixel center, letting clipped endpoints land one pixel beyond the raster.
Cohen sutherland pixel window: window uses pixel count · case 04
The window maximum uses the pixel count rather than the last pixel center, letting clipped endpoints land one pixel beyond the raster.
Cohen sutherland pixel window: window uses pixel count · case 05
The window maximum uses the pixel count rather than the last pixel center, letting clipped endpoints land one pixel beyond the raster.
Cohen sutherland pixel window: iteration budget · case 01
The clip loop assumes one clip per endpoint, so corner-crossing segments needing two clips at each end are reported unresolved.
Cohen sutherland pixel window: iteration budget · case 02
The clip loop assumes one clip per endpoint, so corner-crossing segments needing two clips at each end are reported unresolved.
Cohen sutherland pixel window: iteration budget · case 03
The clip loop assumes one clip per endpoint, so corner-crossing segments needing two clips at each end are reported unresolved.
Cohen sutherland pixel window: iteration budget · case 04
The clip loop assumes one clip per endpoint, so corner-crossing segments needing two clips at each end are reported unresolved.
Cohen sutherland pixel window: iteration budget · case 05
The clip loop assumes one clip per endpoint, so corner-crossing segments needing two clips at each end are reported unresolved.
Liang barsky parametric rect: rect right unclamped · case 01
The clip rectangle is not intersected with the raster on the right, so parameters extend past the last pixel column.
Liang barsky parametric rect: rect right unclamped · case 02
The clip rectangle is not intersected with the raster on the right, so parameters extend past the last pixel column.
Liang barsky parametric rect: rect right unclamped · case 03
The clip rectangle is not intersected with the raster on the right, so parameters extend past the last pixel column.
Liang barsky parametric rect: rect right unclamped · case 04
The clip rectangle is not intersected with the raster on the right, so parameters extend past the last pixel column.
Liang barsky parametric rect: rect right unclamped · case 05
The clip rectangle is not intersected with the raster on the right, so parameters extend past the last pixel column.
Liang barsky parametric rect: entry classified by distance · case 01
Boundaries are classified as entering or leaving by the sign of the boundary distance instead of the direction term, so interior starts treat every plane as leaving.
Liang barsky parametric rect: entry classified by distance · case 02
Boundaries are classified as entering or leaving by the sign of the boundary distance instead of the direction term, so interior starts treat every plane as leaving.
Liang barsky parametric rect: entry classified by distance · case 03
Boundaries are classified as entering or leaving by the sign of the boundary distance instead of the direction term, so interior starts treat every plane as leaving.
Liang barsky parametric rect: entry classified by distance · case 04
Boundaries are classified as entering or leaving by the sign of the boundary distance instead of the direction term, so interior starts treat every plane as leaving.
Liang barsky parametric rect: entry classified by distance · case 05
Boundaries are classified as entering or leaving by the sign of the boundary distance instead of the direction term, so interior starts treat every plane as leaving.
Liang barsky parametric rect: parallel boundary exclusive · case 01
A segment parallel to and lying exactly on a clip boundary is rejected although boundary points are inside.
Liang barsky parametric rect: parallel boundary exclusive · case 02
A segment parallel to and lying exactly on a clip boundary is rejected although boundary points are inside.
Liang barsky parametric rect: parallel boundary exclusive · case 03
A segment parallel to and lying exactly on a clip boundary is rejected although boundary points are inside.
Liang barsky parametric rect: parallel boundary exclusive · case 04
A segment parallel to and lying exactly on a clip boundary is rejected although boundary points are inside.
Liang barsky parametric rect: parallel boundary exclusive · case 05
A segment parallel to and lying exactly on a clip boundary is rejected although boundary points are inside.
Liang barsky parametric rect: entering parameter overwrite · case 01
Each entering boundary overwrites t0, so a later less restrictive entering plane loosens the start parameter.
Liang barsky parametric rect: entering parameter overwrite · case 02
Each entering boundary overwrites t0, so a later less restrictive entering plane loosens the start parameter.
Liang barsky parametric rect: entering parameter overwrite · case 03
Each entering boundary overwrites t0, so a later less restrictive entering plane loosens the start parameter.
Liang barsky parametric rect: entering parameter overwrite · case 04
Each entering boundary overwrites t0, so a later less restrictive entering plane loosens the start parameter.
Liang barsky parametric rect: entering parameter overwrite · case 05
Each entering boundary overwrites t0, so a later less restrictive entering plane loosens the start parameter.
Liang barsky parametric rect: left plane sign · case 01
The left boundary distance has the wrong sign, so segments starting inside acquire a spurious entering parameter.
Liang barsky parametric rect: left plane sign · case 02
The left boundary distance has the wrong sign, so segments starting inside acquire a spurious entering parameter.
Liang barsky parametric rect: left plane sign · case 03
The left boundary distance has the wrong sign, so segments starting inside acquire a spurious entering parameter.
Liang barsky parametric rect: left plane sign · case 04
The left boundary distance has the wrong sign, so segments starting inside acquire a spurious entering parameter.
Liang barsky parametric rect: left plane sign · case 05
The left boundary distance has the wrong sign, so segments starting inside acquire a spurious entering parameter.
Liang barsky parametric rect: integer parameter floor · case 01
Boundary parameters are floor-divided, collapsing fractional intersections to integer parameters.
Liang barsky parametric rect: integer parameter floor · case 02
Boundary parameters are floor-divided, collapsing fractional intersections to integer parameters.
Liang barsky parametric rect: integer parameter floor · case 03
Boundary parameters are floor-divided, collapsing fractional intersections to integer parameters.
Liang barsky parametric rect: integer parameter floor · case 04
Boundary parameters are floor-divided, collapsing fractional intersections to integer parameters.
Liang barsky parametric rect: integer parameter floor · case 05
Boundary parameters are floor-divided, collapsing fractional intersections to integer parameters.
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 ↗