Abstract
A method is disclosed for proving that a restructuring of a browser-rendered client application preserves its appearance and behaviour, by reducing "nothing changed" to a zero-difference exit code whose ability to fail is itself tested. A headless capture renders every view into a deterministic artifact set: the clock and pseudo-random source are replaced before application code runs, periodic timers are recorded rather than scheduled, animations are pinned to a fixed frame, each structured snapshot waits for network idle and DOM quiescence together, and endpoint data is replayed from a frozen snapshot. A comparator diffs baseline and candidate captures across appearance, markup, behaviour, program-surface and data areas. Residual nondeterminism in the structured artifacts is not absorbed by a threshold: a stability measurer captures the unchanged build repeatedly with screenshots off, diffs every run against the first with masking disabled, and writes the exact finding-and-detail keys that varied as a mask stored beside the baseline. Screenshots alone are classified against a fixed pixel noise floor, sub-floor differences printed as notes. The comparator reads the mask from the baseline side only, subtracts matching detail lines individually, drops a finding only when all its lines are masked, and in its full report prints each mask hit with its count, so the mask cannot widen unnoticed. A self-test seeds text-anchored mutations that must each match exactly once, and passes only when the comparator reports each one in its expected area; an anchor that fails to apply is a failure of the oracle.
Creative Commons License

This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Assuncao, gustavo matthew, "Empirically-Masked Deterministic UI Parity Oracle: N-Run Instability Mask Stored With the Baseline, Per-Line Mask Subtraction With Printed Hits, and an Anchored-Mutation Self-Test", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/13226