FA-79053 / Image orientation metadata / Member archive

Displayed video size ignores quarter turns · case 03

Rotated videos are laid out in a landscape box and letterboxed.

Member previewVariant 3 · 3 implementations · 8 checks per implementation

Case contract

Input [matrix, width, height] where matrix is the [a, b, c, d] part of a track display matrix stored as unsigned 32-bit 16.16 fixed-point words (x' = a*x + c*y, y' = b*x + d*y, screen y down). Only signed permutation matrices of unit entries are supported, otherwise "unsupported". A negative determinant means a left-right mirror applied before the rotation; the clockwise rotation is read from the image of the x axis after removing the mirror. Return rotate (0/90/180/270), mirror and the displayed [width, height].

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.

One recorded failure

Sample boundary fixture

This sample comes from the broken implementation of a controlled reproducer.

Boundary fixture	Actual	Expected	Outcome
regression: displayed size swap 0	`{"display": [3412, 3662], "mirror": false, "rotate": 90}`	`{"display": [3662, 3412], "mirror": false, "rotate": 90}`	Failed

MEMBER ARCHIVE

The complete case is available to members.

This record includes three runnable implementations, regression fixtures, execution results, and source hashes.

Member access is invitation-based. Sign in with your invited account to inspect the sources.