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
Encoder flushes all subnormals · case 01
Encoder flushes all subnormals.
Encoder flushes all subnormals · case 02
Encoder flushes all subnormals.
Encoder flushes all subnormals · case 03
Encoder flushes all subnormals.
Encoder flushes all subnormals · case 04
Encoder flushes all subnormals.
Encoder flushes all subnormals · case 05
Encoder flushes all subnormals.
Subnormal rounding carry is masked away · case 01
Subnormal rounding carry is masked away.
Subnormal rounding carry is masked away · case 02
Subnormal rounding carry is masked away.
Subnormal rounding carry is masked away · case 03
Subnormal rounding carry is masked away.
Subnormal rounding carry is masked away · case 04
Subnormal rounding carry is masked away.
Subnormal rounding carry is masked away · case 05
Subnormal rounding carry is masked away.
Encoder treats frexp mantissa as having a unit leading bit · case 01
Encoder treats frexp mantissa as having a unit leading bit.
Encoder treats frexp mantissa as having a unit leading bit · case 02
Encoder treats frexp mantissa as having a unit leading bit.
Encoder treats frexp mantissa as having a unit leading bit · case 03
Encoder treats frexp mantissa as having a unit leading bit.
Encoder treats frexp mantissa as having a unit leading bit · case 04
Encoder treats frexp mantissa as having a unit leading bit.
Encoder treats frexp mantissa as having a unit leading bit · case 05
Encoder treats frexp mantissa as having a unit leading bit.
Normalized significand is rounded before scaling · case 01
Normalized significand is rounded before scaling.
Normalized significand is rounded before scaling · case 02
Normalized significand is rounded before scaling.
Normalized significand is rounded before scaling · case 03
Normalized significand is rounded before scaling.
Normalized significand is rounded before scaling · case 04
Normalized significand is rounded before scaling.
Normalized significand is rounded before scaling · case 05
Normalized significand is rounded before scaling.
Mantissa carry is reset to zero rather than the hidden one · case 01
Mantissa carry is reset to zero rather than the hidden one.
Mantissa carry is reset to zero rather than the hidden one · case 02
Mantissa carry is reset to zero rather than the hidden one.
Mantissa carry is reset to zero rather than the hidden one · case 03
Mantissa carry is reset to zero rather than the hidden one.
Mantissa carry is reset to zero rather than the hidden one · case 04
Mantissa carry is reset to zero rather than the hidden one.
Mantissa carry is reset to zero rather than the hidden one · case 05
Mantissa carry is reset to zero rather than the hidden one.
Mantissa carry does not advance the exponent · case 01
Mantissa carry does not advance the exponent.
Mantissa carry does not advance the exponent · case 02
Mantissa carry does not advance the exponent.
Mantissa carry does not advance the exponent · case 03
Mantissa carry does not advance the exponent.
Mantissa carry does not advance the exponent · case 04
Mantissa carry does not advance the exponent.
Mantissa carry does not advance the exponent · case 05
Mantissa carry does not advance the exponent.
Encoding applies the bias in the wrong direction · case 01
Encoding applies the bias in the wrong direction.
Encoding applies the bias in the wrong direction · case 02
Encoding applies the bias in the wrong direction.
Encoding applies the bias in the wrong direction · case 03
Encoding applies the bias in the wrong direction.
Encoding applies the bias in the wrong direction · case 04
Encoding applies the bias in the wrong direction.
Encoding applies the bias in the wrong direction · case 05
Encoding applies the bias in the wrong direction.
The encoded fraction retains its implicit leading one · case 01
The encoded fraction retains its implicit leading one.
The encoded fraction retains its implicit leading one · case 02
The encoded fraction retains its implicit leading one.
The encoded fraction retains its implicit leading one · case 03
The encoded fraction retains its implicit leading one.
The encoded fraction retains its implicit leading one · case 04
The encoded fraction retains its implicit leading one.
The encoded fraction retains its implicit leading one · case 05
The encoded fraction retains its implicit leading one.
Exponent and fraction are intersected instead of combined · case 01
Exponent and fraction are intersected instead of combined.
Exponent and fraction are intersected instead of combined · case 02
Exponent and fraction are intersected instead of combined.
Exponent and fraction are intersected instead of combined · case 03
Exponent and fraction are intersected instead of combined.
Exponent and fraction are intersected instead of combined · case 04
Exponent and fraction are intersected instead of combined.
Exponent and fraction are intersected instead of combined · case 05
Exponent and fraction are intersected instead of combined.
Exponent adjustment canonicalizes negative zero · case 01
Exponent adjustment canonicalizes negative zero.
Exponent adjustment canonicalizes negative zero · case 02
Exponent adjustment canonicalizes negative zero.
Exponent adjustment canonicalizes negative zero · case 03
Exponent adjustment canonicalizes negative zero.
Exponent adjustment canonicalizes negative zero · case 04
Exponent adjustment canonicalizes negative zero.
Exponent adjustment canonicalizes negative zero · case 05
Exponent adjustment canonicalizes negative zero.
Exponent adjustment reverses the requested shift · case 01
Exponent adjustment reverses the requested shift.
Exponent adjustment reverses the requested shift · case 02
Exponent adjustment reverses the requested shift.
Exponent adjustment reverses the requested shift · case 03
Exponent adjustment reverses the requested shift.
Exponent adjustment reverses the requested shift · case 04
Exponent adjustment reverses the requested shift.
Exponent adjustment reverses the requested shift · case 05
Exponent adjustment reverses the requested shift.
Exponent adjustment discards the mantissa sign · case 01
Exponent adjustment discards the mantissa sign.
Exponent adjustment discards the mantissa sign · case 02
Exponent adjustment discards the mantissa sign.
Exponent adjustment discards the mantissa sign · case 03
Exponent adjustment discards the mantissa sign.
Exponent adjustment discards the mantissa sign · case 04
Exponent adjustment discards the mantissa sign.
Exponent adjustment discards the mantissa sign · case 05
Exponent adjustment discards the mantissa sign.
Exponent adjustment rejects the highest finite binade · case 01
Exponent adjustment rejects the highest finite binade.
Exponent adjustment rejects the highest finite binade · case 02
Exponent adjustment rejects the highest finite binade.
Exponent adjustment rejects the highest finite binade · case 03
Exponent adjustment rejects the highest finite binade.
Exponent adjustment rejects the highest finite binade · case 04
Exponent adjustment rejects the highest finite binade.
Exponent adjustment rejects the highest finite binade · case 05
Exponent adjustment rejects the highest finite binade.
Exponent overflow loses its negative sign · case 01
Exponent overflow loses its negative sign.
Exponent overflow loses its negative sign · case 02
Exponent overflow loses its negative sign.
Exponent overflow loses its negative sign · case 03
Exponent overflow loses its negative sign.
Exponent overflow loses its negative sign · case 04
Exponent overflow loses its negative sign.
Exponent overflow loses its negative sign · case 05
Exponent overflow loses its negative sign.
Exponent adjustment flushes representable subnormals · case 01
Exponent adjustment flushes representable subnormals.
Exponent adjustment flushes representable subnormals · case 02
Exponent adjustment flushes representable subnormals.
Exponent adjustment flushes representable subnormals · case 03
Exponent adjustment flushes representable subnormals.
Exponent adjustment flushes representable subnormals · case 04
Exponent adjustment flushes representable subnormals.
Exponent adjustment flushes representable subnormals · case 05
Exponent adjustment flushes representable subnormals.
Exponent adjustment strips the sign after underflow · case 01
Exponent adjustment strips the sign after underflow.
Exponent adjustment strips the sign after underflow · case 02
Exponent adjustment strips the sign after underflow.
Exponent adjustment strips the sign after underflow · case 03
Exponent adjustment strips the sign after underflow.
Exponent adjustment strips the sign after underflow · case 04
Exponent adjustment strips the sign after underflow.
Exponent adjustment strips the sign after underflow · case 05
Exponent adjustment strips the sign after underflow.
ULP distance validates only its right operand · case 01
ULP distance validates only its right operand.
ULP distance validates only its right operand · case 02
ULP distance validates only its right operand.
ULP distance validates only its right operand · case 03
ULP distance validates only its right operand.
ULP distance validates only its right operand · case 04
ULP distance validates only its right operand.
ULP distance validates only its right operand · case 05
ULP distance validates only its right operand.
ULP distance validates only its left operand · case 01
ULP distance validates only its left operand.
ULP distance validates only its left operand · case 02
ULP distance validates only its left operand.
ULP distance validates only its left operand · case 03
ULP distance validates only its left operand.
ULP distance validates only its left operand · case 04
ULP distance validates only its left operand.
ULP distance validates only its left operand · case 05
ULP distance validates only its left operand.
ULP distance leaves the first negative encoding unsigned · case 01
ULP distance leaves the first negative encoding unsigned.
ULP distance leaves the first negative encoding unsigned · case 02
ULP distance leaves the first negative encoding unsigned.
ULP distance leaves the first negative encoding unsigned · case 03
ULP distance leaves the first negative encoding unsigned.
ULP distance leaves the first negative encoding unsigned · case 04
ULP distance leaves the first negative encoding unsigned.
ULP distance leaves the first negative encoding unsigned · case 05
ULP distance leaves the first negative encoding unsigned.
ULP distance complements the second negative encoding without zero alignment · case 01
ULP distance complements the second negative encoding without zero alignment.
ULP distance complements the second negative encoding without zero alignment · case 02
ULP distance complements the second negative encoding without zero alignment.
ULP distance complements the second negative encoding without zero alignment · case 03
ULP distance complements the second negative encoding without zero alignment.
ULP distance complements the second negative encoding without zero alignment · case 04
ULP distance complements the second negative encoding without zero alignment.
ULP distance complements the second negative encoding without zero alignment · case 05
ULP distance complements the second negative encoding without zero alignment.
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 ↗