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
Weak type resolution: terminator run extent · case 01
Only the terminator touching the number is attached.
Weak type resolution: terminator run extent · case 02
Only the terminator touching the number is attached.
Weak type resolution: terminator run extent · case 03
Only the terminator touching the number is attached.
Weak type resolution: terminator run extent · case 04
Only the terminator touching the number is attached.
Weak type resolution: terminator run extent · case 05
Only the terminator touching the number is attached.
Weak type resolution: leftover separator neutralisation · case 01
Unattached currency signs keep a weak type and block neutral resolution.
Weak type resolution: leftover separator neutralisation · case 02
Unattached currency signs keep a weak type and block neutral resolution.
Weak type resolution: leftover separator neutralisation · case 03
Unattached currency signs keep a weak type and block neutral resolution.
Weak type resolution: leftover separator neutralisation · case 04
Unattached currency signs keep a weak type and block neutral resolution.
Weak type resolution: leftover separator neutralisation · case 05
Unattached currency signs keep a weak type and block neutral resolution.
Weak type resolution: European number context start · case 01
Numbers at the start of a sequence ignore the sequence start type.
Weak type resolution: European number context start · case 02
Numbers at the start of a sequence ignore the sequence start type.
Weak type resolution: European number context start · case 03
Numbers at the start of a sequence ignore the sequence start type.
Weak type resolution: European number context start · case 04
Numbers at the start of a sequence ignore the sequence start type.
Weak type resolution: European number context start · case 05
Numbers at the start of a sequence ignore the sequence start type.
Neutral run resolution: numbers as right-to-left context · case 01
Neutrals between Hebrew text and a number fall back to the embedding direction.
Neutral run resolution: numbers as right-to-left context · case 02
Neutrals between Hebrew text and a number fall back to the embedding direction.
Neutral run resolution: numbers as right-to-left context · case 03
Neutrals between Hebrew text and a number fall back to the embedding direction.
Neutral run resolution: numbers as right-to-left context · case 04
Neutrals between Hebrew text and a number fall back to the embedding direction.
Neutral run resolution: numbers as right-to-left context · case 05
Neutrals between Hebrew text and a number fall back to the embedding direction.
Neutral run resolution: embedding direction parity · case 01
Unresolvable neutrals take the opposite of the embedding direction.
Neutral run resolution: embedding direction parity · case 02
Unresolvable neutrals take the opposite of the embedding direction.
Neutral run resolution: embedding direction parity · case 03
Unresolvable neutrals take the opposite of the embedding direction.
Neutral run resolution: embedding direction parity · case 04
Unresolvable neutrals take the opposite of the embedding direction.
Neutral run resolution: embedding direction parity · case 05
Unresolvable neutrals take the opposite of the embedding direction.
Neutral run resolution: start-of-sequence context · case 01
Leading neutrals ignore the sequence start type.
Neutral run resolution: start-of-sequence context · case 02
Leading neutrals ignore the sequence start type.
Neutral run resolution: start-of-sequence context · case 03
Leading neutrals ignore the sequence start type.
Neutral run resolution: start-of-sequence context · case 04
Leading neutrals ignore the sequence start type.
Neutral run resolution: start-of-sequence context · case 05
Leading neutrals ignore the sequence start type.
Neutral run resolution: end-of-sequence context · case 01
Trailing neutrals resolve against the wrong edge.
Neutral run resolution: end-of-sequence context · case 02
Trailing neutrals resolve against the wrong edge.
Neutral run resolution: end-of-sequence context · case 03
Trailing neutrals resolve against the wrong edge.
Neutral run resolution: end-of-sequence context · case 04
Trailing neutrals resolve against the wrong edge.
Neutral run resolution: end-of-sequence context · case 05
Trailing neutrals resolve against the wrong edge.
Neutral run resolution: neutral run grouping · case 01
Neutral runs resolve piecemeal with inconsistent directions.
Neutral run resolution: neutral run grouping · case 02
Neutral runs resolve piecemeal with inconsistent directions.
Neutral run resolution: neutral run grouping · case 03
Neutral runs resolve piecemeal with inconsistent directions.
Neutral run resolution: neutral run grouping · case 04
Neutral runs resolve piecemeal with inconsistent directions.
Neutral run resolution: neutral run grouping · case 05
Neutral runs resolve piecemeal with inconsistent directions.
Neutral run resolution: segment separator neutrality · case 01
Tabs split neutral runs and keep a separator class.
Neutral run resolution: segment separator neutrality · case 02
Tabs split neutral runs and keep a separator class.
Neutral run resolution: segment separator neutrality · case 03
Tabs split neutral runs and keep a separator class.
Neutral run resolution: segment separator neutrality · case 04
Tabs split neutral runs and keep a separator class.
Neutral run resolution: segment separator neutrality · case 05
Tabs split neutral runs and keep a separator class.
Implicit level runs: number raise at even level · case 01
Numbers in LTR text collapse into the surrounding RTL run.
Implicit level runs: number raise at even level · case 02
Numbers in LTR text collapse into the surrounding RTL run.
Implicit level runs: number raise at even level · case 03
Numbers in LTR text collapse into the surrounding RTL run.
Implicit level runs: number raise at even level · case 04
Numbers in LTR text collapse into the surrounding RTL run.
Implicit level runs: number raise at even level · case 05
Numbers in LTR text collapse into the surrounding RTL run.
Implicit level runs: numbers at odd level · case 01
Numbers inside RTL text read right-to-left digit by digit.
Implicit level runs: numbers at odd level · case 02
Numbers inside RTL text read right-to-left digit by digit.
Implicit level runs: numbers at odd level · case 03
Numbers inside RTL text read right-to-left digit by digit.
Implicit level runs: numbers at odd level · case 04
Numbers inside RTL text read right-to-left digit by digit.
Implicit level runs: numbers at odd level · case 05
Numbers inside RTL text read right-to-left digit by digit.
Implicit level runs: level parity test · case 01
Deep even embeddings are treated like odd ones.
Implicit level runs: level parity test · case 02
Deep even embeddings are treated like odd ones.
Implicit level runs: level parity test · case 03
Deep even embeddings are treated like odd ones.
Implicit level runs: level parity test · case 04
Deep even embeddings are treated like odd ones.
Implicit level runs: level parity test · case 05
Deep even embeddings are treated like odd ones.
Implicit level runs: run end bookkeeping · case 01
Level runs report ends one short.
Implicit level runs: run end bookkeeping · case 02
Level runs report ends one short.
Implicit level runs: run end bookkeeping · case 03
Level runs report ends one short.
Implicit level runs: run end bookkeeping · case 04
Level runs report ends one short.
Implicit level runs: run end bookkeeping · case 05
Level runs report ends one short.
Implicit level runs: run level equality · case 01
Runs at levels 1 and 3 are merged.
Implicit level runs: run level equality · case 02
Runs at levels 1 and 3 are merged.
Implicit level runs: run level equality · case 03
Runs at levels 1 and 3 are merged.
Implicit level runs: run level equality · case 04
Runs at levels 1 and 3 are merged.
Implicit level runs: run level equality · case 05
Runs at levels 1 and 3 are merged.
Visual reordering by levels: highest level start · case 01
Numbers embedded at an even level inside RTL text read backwards.
Visual reordering by levels: highest level start · case 02
Numbers embedded at an even level inside RTL text read backwards.
Visual reordering by levels: highest level start · case 03
Numbers embedded at an even level inside RTL text read backwards.
Visual reordering by levels: highest level start · case 04
Numbers embedded at an even level inside RTL text read backwards.
Visual reordering by levels: highest level start · case 05
Numbers embedded at an even level inside RTL text read backwards.
Visual reordering by levels: reversal loop bound · case 01
Runs at the lowest odd level stay in logical order.
Visual reordering by levels: reversal loop bound · case 02
Runs at the lowest odd level stay in logical order.
Visual reordering by levels: reversal loop bound · case 03
Runs at the lowest odd level stay in logical order.
Visual reordering by levels: reversal loop bound · case 04
Runs at the lowest odd level stay in logical order.
Visual reordering by levels: reversal loop bound · case 05
Runs at the lowest odd level stay in logical order.
Visual reordering by levels: pass step · case 01
Levels of one parity are never reversed as their own pass.
Visual reordering by levels: pass step · case 02
Levels of one parity are never reversed as their own pass.
Visual reordering by levels: pass step · case 03
Levels of one parity are never reversed as their own pass.
Visual reordering by levels: pass step · case 04
Levels of one parity are never reversed as their own pass.
Visual reordering by levels: pass step · case 05
Levels of one parity are never reversed as their own pass.
Visual reordering by levels: no odd level shortcut · case 01
Pure LTR lines with embedded even levels are reversed.
Visual reordering by levels: no odd level shortcut · case 02
Pure LTR lines with embedded even levels are reversed.
Visual reordering by levels: no odd level shortcut · case 03
Pure LTR lines with embedded even levels are reversed.
Visual reordering by levels: no odd level shortcut · case 04
Pure LTR lines with embedded even levels are reversed.
Visual reordering by levels: no odd level shortcut · case 05
Pure LTR lines with embedded even levels are reversed.
Visual reordering by levels: visual map direction · case 01
Callers receive a logical-to-visual map instead of visual order.
Visual reordering by levels: visual map direction · case 02
Callers receive a logical-to-visual map instead of visual order.
Visual reordering by levels: visual map direction · case 03
Callers receive a logical-to-visual map instead of visual order.
Visual reordering by levels: visual map direction · case 04
Callers receive a logical-to-visual map instead of visual order.
Visual reordering by levels: visual map direction · case 05
Callers receive a logical-to-visual map instead of visual order.
Line-end whitespace level reset: isolate controls in reset runs · case 01
Isolate controls at the line end keep embedded levels.
Line-end whitespace level reset: isolate controls in reset runs · case 02
Isolate controls at the line end keep embedded levels.
Line-end whitespace level reset: isolate controls in reset runs · case 03
Isolate controls at the line end keep embedded levels.
Line-end whitespace level reset: isolate controls in reset runs · case 04
Isolate controls at the line end keep embedded levels.
Line-end whitespace level reset: isolate controls in reset runs · case 05
Isolate controls at the line end keep embedded levels.
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 ↗