FA-79076 / Image orientation metadata / Open access

Lossless transpose keeps stored dimensions · case 01

Tag 5 and 7 JPEGs are reported with stored width and height after a lossless bake.

Verified by executionVariant 1 · 8 checks per implementationDownload source bundle ↓JSON ↗

ROOT CAUSE

Only the pure rotations swap the reported size.

VERIFIED REPAIR

Swap the trimmed dimensions whenever the decomposition includes a transpose.

Unsuccessful approach: Requiring a horizontal flip as well misses tags 5 and 6.

Case contract

Input [w, h, tag, sampling]. A lossless JPEG transform that bakes the orientation can only move whole MCUs (444: 8x8, 422: 16 wide x 8 high, 420: 16x16, 440: 8 wide x 16 high), so any partial MCU column or row that would move to the leading edge is trimmed. The tag decomposes into stored-space flips (fx, fy) followed by an optional transpose; fx trims the width and fy trims the height to a multiple of the MCU size. Invalid tags are treated as 1. Return the displayed [width, height], "refuse" when a trimmed side becomes 0, or "unknown-sampling".

Why this case matters

Camera, phone and scanner images carry an orientation hint separately from the stored pixels; galleries, thumbnailers, editors and upload pipelines must interpret it consistently or photos appear sideways, mirrored or doubly rotated.

1 / The failure

Exit 1

"""Failure Map reference implementation. Python standard library only."""
import json

N = 1
observations = []
def solve(x):
    w, h, tag, sampling = x
    mcu = {'444': (8, 8), '422': (16, 8), '420': (16, 16), '440': (8, 16)}
    if sampling not in mcu:
        return 'unknown-sampling'
    mw, mh = mcu[sampling]
    flips = {1: (0, 0, 0), 2: (1, 0, 0), 3: (1, 1, 0), 4: (0, 1, 0), 5: (0, 0, 1), 6: (0, 1, 1), 7: (1, 1, 1), 8: (1, 0, 1)}
    fx, fy, tr = flips.get(tag, (0, 0, 0))
    tw = w - w % mw if fx else w
    th = h - h % mh if fy else h
    if tw == 0 or th == 0:
        return 'refuse'
    if tag in (6, 8):
        tw, th = th, tw
    return [tw, th]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
fixtures = [[[[24, 16, 5, '422'], [16, 24]], [[24, 20, 5, '420'], [20, 24]], [[24, 12, 0, '440'], [24, 12]], [[2969, 96, 1, '422'], [2969, 96]], [[40, 2600, 0, '440'], [40, 2600]], [[24, 20, 0, '444'], [24, 20]], [[16, 24, 0, '420'], [16, 24]], [[12, 16, 5, '444'], [16, 12]]], [[[2388, 13, 5, '420'], [13, 2388]], [[16, 20, 5, '440'], [20, 16]], [[12, 1946, 4, '420'], [12, 1936]], [[8, 12, 1, '411'], 'unknown-sampling'], [[25, 8, 8, '420'], [8, 16]], [[8, 2948, 7, '422'], 'refuse'], [[24, 24, 2, '440'], [24, 24]], [[2550, 16, 5, '444'], [16, 2550]]], [[[8, 12, 5, '422'], [12, 8]], [[8, 20, 6, '420'], [16, 8]], [[40, 24, 5, '422'], [24, 40]], [[24, 8, 4, '444'], [24, 8]], [[12, 12, 3, '420'], 'refuse'], [[24, 24, 0, '420'], [24, 24]], [[40, 24, 0, '422'], [40, 24]], [[40, 8, 7, '444'], [8, 40]]], [[[40, 81, 5, '444'], [81, 40]], [[24, 16, 6, '440'], [16, 24]], [[12, 32, 5, '420'], [32, 12]], [[40, 12, 2, '444'], [40, 12]], [[24, 24, 4, '420'], [24, 16]], [[24, 4, 3, '422'], 'refuse'], [[8, 8, 0, '411'], 'unknown-sampling'], [[2236, 16, 7, '422'], [16, 2224]]], [[[8, 24, 5, '422'], [24, 8]], [[1545, 12, 6, '422'], [8, 1545]], [[16, 24, 2, '411'], 'unknown-sampling'], [[40, 18, 1, '420'], [40, 18]], [[40, 8, 1, '440'], [40, 8]], [[16, 8, 8, '422'], [8, 16]], [[1533, 16, 3, '411'], 'unknown-sampling'], [[8, 67, 7, '444'], [64, 8]]]]
labels = ["regression: output dimension swap", "repair trap", "combined fault", "control", "control", "boundary", "boundary", "control"]
for i, (args, expected) in enumerate(fixtures[N-1]):
    check("%s %d" % (labels[i % len(labels)], i), solve(args), expected)
print(json.dumps({"observations": observations, "passed": all(x["passed"] for x in observations)}, ensure_ascii=False))
raise SystemExit(0 if all(x["passed"] for x in observations) else 1)

Boundary fixture	Actual	Expected	Outcome
regression: output dimension swap 0	`[24, 16]`	`[16, 24]`	Failed
repair trap 1	`[24, 20]`	`[20, 24]`	Failed
combined fault 2	`[24, 12]`	`[24, 12]`	Passed
control 3	`[2969, 96]`	`[2969, 96]`	Passed
control 4	`[40, 2600]`	`[40, 2600]`	Passed
boundary 5	`[24, 20]`	`[24, 20]`	Passed
boundary 6	`[16, 24]`	`[16, 24]`	Passed
control 7	`[12, 16]`	`[16, 12]`	Failed

SHA-256 / 3c28c24ad2d79e3122521c07bcc30a87426237ae5a6aeaff6aaf947b17f71127

2 / The unsuccessful fix

Exit 1

"""Failure Map reference implementation. Python standard library only."""
import json

N = 1
observations = []
def solve(x):
    w, h, tag, sampling = x
    mcu = {'444': (8, 8), '422': (16, 8), '420': (16, 16), '440': (8, 16)}
    if sampling not in mcu:
        return 'unknown-sampling'
    mw, mh = mcu[sampling]
    flips = {1: (0, 0, 0), 2: (1, 0, 0), 3: (1, 1, 0), 4: (0, 1, 0), 5: (0, 0, 1), 6: (0, 1, 1), 7: (1, 1, 1), 8: (1, 0, 1)}
    fx, fy, tr = flips.get(tag, (0, 0, 0))
    tw = w - w % mw if fx else w
    th = h - h % mh if fy else h
    if tw == 0 or th == 0:
        return 'refuse'
    if fx and tr:
        tw, th = th, tw
    return [tw, th]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
fixtures = [[[[24, 16, 5, '422'], [16, 24]], [[24, 20, 5, '420'], [20, 24]], [[24, 12, 0, '440'], [24, 12]], [[2969, 96, 1, '422'], [2969, 96]], [[40, 2600, 0, '440'], [40, 2600]], [[24, 20, 0, '444'], [24, 20]], [[16, 24, 0, '420'], [16, 24]], [[12, 16, 5, '444'], [16, 12]]], [[[2388, 13, 5, '420'], [13, 2388]], [[16, 20, 5, '440'], [20, 16]], [[12, 1946, 4, '420'], [12, 1936]], [[8, 12, 1, '411'], 'unknown-sampling'], [[25, 8, 8, '420'], [8, 16]], [[8, 2948, 7, '422'], 'refuse'], [[24, 24, 2, '440'], [24, 24]], [[2550, 16, 5, '444'], [16, 2550]]], [[[8, 12, 5, '422'], [12, 8]], [[8, 20, 6, '420'], [16, 8]], [[40, 24, 5, '422'], [24, 40]], [[24, 8, 4, '444'], [24, 8]], [[12, 12, 3, '420'], 'refuse'], [[24, 24, 0, '420'], [24, 24]], [[40, 24, 0, '422'], [40, 24]], [[40, 8, 7, '444'], [8, 40]]], [[[40, 81, 5, '444'], [81, 40]], [[24, 16, 6, '440'], [16, 24]], [[12, 32, 5, '420'], [32, 12]], [[40, 12, 2, '444'], [40, 12]], [[24, 24, 4, '420'], [24, 16]], [[24, 4, 3, '422'], 'refuse'], [[8, 8, 0, '411'], 'unknown-sampling'], [[2236, 16, 7, '422'], [16, 2224]]], [[[8, 24, 5, '422'], [24, 8]], [[1545, 12, 6, '422'], [8, 1545]], [[16, 24, 2, '411'], 'unknown-sampling'], [[40, 18, 1, '420'], [40, 18]], [[40, 8, 1, '440'], [40, 8]], [[16, 8, 8, '422'], [8, 16]], [[1533, 16, 3, '411'], 'unknown-sampling'], [[8, 67, 7, '444'], [64, 8]]]]
labels = ["regression: output dimension swap", "repair trap", "combined fault", "control", "control", "boundary", "boundary", "control"]
for i, (args, expected) in enumerate(fixtures[N-1]):
    check("%s %d" % (labels[i % len(labels)], i), solve(args), expected)
print(json.dumps({"observations": observations, "passed": all(x["passed"] for x in observations)}, ensure_ascii=False))
raise SystemExit(0 if all(x["passed"] for x in observations) else 1)

Boundary fixture	Actual	Expected	Outcome
regression: output dimension swap 0	`[24, 16]`	`[16, 24]`	Failed
repair trap 1	`[24, 20]`	`[20, 24]`	Failed
combined fault 2	`[24, 12]`	`[24, 12]`	Passed
control 3	`[2969, 96]`	`[2969, 96]`	Passed
control 4	`[40, 2600]`	`[40, 2600]`	Passed
boundary 5	`[24, 20]`	`[24, 20]`	Passed
boundary 6	`[16, 24]`	`[16, 24]`	Passed
control 7	`[12, 16]`	`[16, 12]`	Failed

SHA-256 / a858bd1f750a4e856e0b7e0e6508524be65d2517eedba9d3246d6d8ba0299de9

3 / The verified repair

Exit 0

"""Failure Map reference implementation. Python standard library only."""
import json

N = 1
observations = []
def solve(x):
    w, h, tag, sampling = x
    mcu = {'444': (8, 8), '422': (16, 8), '420': (16, 16), '440': (8, 16)}
    if sampling not in mcu:
        return 'unknown-sampling'
    mw, mh = mcu[sampling]
    flips = {1: (0, 0, 0), 2: (1, 0, 0), 3: (1, 1, 0), 4: (0, 1, 0), 5: (0, 0, 1), 6: (0, 1, 1), 7: (1, 1, 1), 8: (1, 0, 1)}
    fx, fy, tr = flips.get(tag, (0, 0, 0))
    tw = w - w % mw if fx else w
    th = h - h % mh if fy else h
    if tw == 0 or th == 0:
        return 'refuse'
    if tr:
        tw, th = th, tw
    return [tw, th]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
fixtures = [[[[24, 16, 5, '422'], [16, 24]], [[24, 20, 5, '420'], [20, 24]], [[24, 12, 0, '440'], [24, 12]], [[2969, 96, 1, '422'], [2969, 96]], [[40, 2600, 0, '440'], [40, 2600]], [[24, 20, 0, '444'], [24, 20]], [[16, 24, 0, '420'], [16, 24]], [[12, 16, 5, '444'], [16, 12]]], [[[2388, 13, 5, '420'], [13, 2388]], [[16, 20, 5, '440'], [20, 16]], [[12, 1946, 4, '420'], [12, 1936]], [[8, 12, 1, '411'], 'unknown-sampling'], [[25, 8, 8, '420'], [8, 16]], [[8, 2948, 7, '422'], 'refuse'], [[24, 24, 2, '440'], [24, 24]], [[2550, 16, 5, '444'], [16, 2550]]], [[[8, 12, 5, '422'], [12, 8]], [[8, 20, 6, '420'], [16, 8]], [[40, 24, 5, '422'], [24, 40]], [[24, 8, 4, '444'], [24, 8]], [[12, 12, 3, '420'], 'refuse'], [[24, 24, 0, '420'], [24, 24]], [[40, 24, 0, '422'], [40, 24]], [[40, 8, 7, '444'], [8, 40]]], [[[40, 81, 5, '444'], [81, 40]], [[24, 16, 6, '440'], [16, 24]], [[12, 32, 5, '420'], [32, 12]], [[40, 12, 2, '444'], [40, 12]], [[24, 24, 4, '420'], [24, 16]], [[24, 4, 3, '422'], 'refuse'], [[8, 8, 0, '411'], 'unknown-sampling'], [[2236, 16, 7, '422'], [16, 2224]]], [[[8, 24, 5, '422'], [24, 8]], [[1545, 12, 6, '422'], [8, 1545]], [[16, 24, 2, '411'], 'unknown-sampling'], [[40, 18, 1, '420'], [40, 18]], [[40, 8, 1, '440'], [40, 8]], [[16, 8, 8, '422'], [8, 16]], [[1533, 16, 3, '411'], 'unknown-sampling'], [[8, 67, 7, '444'], [64, 8]]]]
labels = ["regression: output dimension swap", "repair trap", "combined fault", "control", "control", "boundary", "boundary", "control"]
for i, (args, expected) in enumerate(fixtures[N-1]):
    check("%s %d" % (labels[i % len(labels)], i), solve(args), expected)
print(json.dumps({"observations": observations, "passed": all(x["passed"] for x in observations)}, ensure_ascii=False))
raise SystemExit(0 if all(x["passed"] for x in observations) else 1)

Boundary fixture	Actual	Expected	Outcome
regression: output dimension swap 0	`[16, 24]`	`[16, 24]`	Passed
repair trap 1	`[20, 24]`	`[20, 24]`	Passed
combined fault 2	`[24, 12]`	`[24, 12]`	Passed
control 3	`[2969, 96]`	`[2969, 96]`	Passed
control 4	`[40, 2600]`	`[40, 2600]`	Passed
boundary 5	`[24, 20]`	`[24, 20]`	Passed
boundary 6	`[16, 24]`	`[16, 24]`	Passed
control 7	`[16, 12]`	`[16, 12]`	Passed

SHA-256 / e25da64c32f4dce8319625cee07c3a3f9f0958f728b6bdb951588ce5e50c1a6b

Verification & scope

A deterministic bounded teaching model with a stipulated contract; it makes no claim of conformance to any published specification. This reproducer isolates one failure mechanism. Results cover the supplied fixtures. Variants within a family share a test contract and should remain grouped when constructing evaluation splits. Related mechanisms with a shared evaluation_group must also remain together; these controlled models are not independent production incidents.

Observations recorded using Python 3.12.14 at 2026-09-29T14:49:41.159349+00:00.

Case digest / 0f66c5547159019d257da574dbefa3b3b831e89b010e6b16fa5257419cf54a5b