9.1 KiB
Native Saved A spacing and reopen: b10, Runs 4 and 5
Closed native evidence for source b10b28cefd1d0794073ba241a0c8de9a83e85f68. The parent operated the frozen GUI; this archive was prepared by reading the closed records, independently comparing the saved Project outputs, and inspecting before/after images. No program was run and no native input was sent during archiving.
Observed edits and independently checked scope
Four intent JSONs identify Saved A 70200000-0000-4000-8000-000000000040, target, field, direction and requested drag before the corresponding observations/results. Their timestamps precede the before/after capture records and comparator execution. The operator notes report that they were written before each drag and before reading the saved result; the archive preserves that provenance rather than claiming an independent temporal recording of the input itself.
| Edit | Displayed zoom | Requested window coordinates | Observed saved coordinate | Independent acceptance |
|---|---|---|---|---|
| Updated occurrence 021, message y | 100% | (650,477)→(650,507) | 252→282 | Only selected message coordinate changes; finite, nonzero increase |
| Assessment lifeline 012, lane x | 100% | (871,395)→(911,395) | 640→680 | Only selected lane coordinate changes; finite, nonzero increase |
| Updated occurrence 021, message y | 51% | (650,613)→(650,593) | 282→242.9375 | Only selected message coordinate changes; finite, nonzero decrease |
| Assessment lifeline 012, lane x | 51% | (794,495)→(764,495) | 680→621.40625 | Only selected lane coordinate changes; finite, nonzero decrease |
The resulting coordinates and deltas are observations, not independently exact expected endpoints. Rounded zoom labels and automation coordinates do not establish the application's actual logical float input bits. The four scoped checks explicitly report independently_exact_final_coordinate: false.
Each scoped check passes complete remaining-Project equality. Archive preparation additionally compared the complete canonical baseline and actual JSON: each differs only in the one numeric token at A's messages/1/send_y or lanes/1/x. The target identity was checked against that record's direct occurrence address or lifeline ID and the recorded intent. All other canonical bytes remain exact, including scalar representations and ordered arrays. Thus the edits preserve semantic order, IDs, other messages/lanes, Saved B, C's earlier repair and unfinished position, names, relationships and all other Project values. This proves scope and direction, not native motion magnitude or continuous tracking. The four message, lane, smaller-zoom message and smaller-zoom lane outputs preserve exact observations and guards.
Before/after images show the moved second arrow while the first arrow and lanes stay in place, and the moved right lane with attached endpoints following while other arrow heights stay in place. All four pairs were inspected during archiving: observations 01/02, 05/06, 09/10 and 13/14. At the smaller zoom, the reported arrow landing is approximately y594 after the requested y593 endpoint; no exact-pixel assertion is substituted for that observation. A background pan before the first lane edit revealed clipped headers. Three minus clicks produced displayed 100→80→64→51%. These are separate camera actions, not model edits during a drag. See Run 4 operator notes.
Exact restoration and durability
The four edits are cumulative. Run 4 begins with the earlier C repair only: checkpoint 01 SHA-256 fc1e12f18de4f3841140998657a8e117f41711ec80ec0bc6aebf0e9c0d61f827. After each edit the operator saved, performed one model Undo and saved, then one model Redo and saved. The eight same comparisons verify exact complete saved-Project restoration; checkpoint source bytes also match each respective baseline or accepted file. Their native Undo/Redo states are documented separately in the images and notes, not inferred from comparator execution.
| Edit | Baseline | Accepted | Undo equals baseline | Redo equals accepted |
|---|---|---|---|---|
| Message 100% | 01 | 02 | 03 | 04 |
| Lane 100% | 05 (=04) | 06 | 07 | 08 |
| Message 51% | 09 (=08) | 10 | 11 | 12 |
| Lane 51% | 13 (=12) | 14 | 15 | 16 |
Run 4 PID 9180 closed normally by observed Alt+F4, exit 0, at 15:38:32 UTC. Run 5 PID 8700 launched through the same recipe after checking that close, reopened the exact incident file, and explicitly selected Saved A. The reopen image shows the fitted arrangement at displayed 74%. Last-view or camera persistence is not claimed. Save all and the ninth exact reopen comparison preserve the entire accepted Project. Run 5 closed normally by observed Alt+F4, exit 0, at 15:41:29 UTC. Both runs have empty stdout/stderr and unchanged bound immutable inputs.
Run 4 checkpoint 16 and Run 5 checkpoint 01 are exactly 6,217 bytes, SHA-256 b746965f034a4befb449d17e568dd5ae0b2ee81b3c34f82883e48e0c54833b25. All thirteen comparisons (four scoped, nine exact) report exit 0, unchanged inputs and acceptance. Every bound output size/hash, before/after capture pair and checkpoint source binding was rechecked during archive preparation. No comparator was rerun.
Capture and scope limits
Sixteen Run 4 JPEGs and one Run 5 JPEG have 17 distinct SHA-256 values matching their metadata and JPEG signatures/MIME. Nine images were visually inspected: the four before/after pairs and reopened A. Hash distinctness alone does not validate every operator description. Original notes and metadata remain byte exact, including compressed wording.
The available drag API supplies start/end only; no held midpoint was captured. These records cannot establish temporal tracking, latency, intermediate jumps or absence of a jump specifically on release. Native exact pointer bits, subpixel endpoint accuracy, broad accessibility, smaller-window Fit, all lifecycle cases, definition navigation and service edits remain outside these runs. The complementary event/controller/mounted tests are separate evidence, not substitute native observations. No full native-plan or release acceptance follows from this archive.
Published prerequisites and integrity
All three current prerequisite topics are pinned to 7cf1d3cbaa40726731997e9de3c4737c3477c270:
- Scoped-motion comparator build and admission evidence, including original exact modes, synthetic refusal checks, source/dependency captures and transitive GUI/gate links.
- Runs 2/3 C repair and reopen, supplying Run 4's baseline and preceding normal-close chain. Its earlier Run 1 link supplies the immutable package recipe, fixture/plan, manifest and launch script.
- Complementary native spacing evaluation, separating exact event-boundary tests, scoped native checks and unobserved temporal behavior. The original immutable plan remains historical evidence.
The bound motion comparator is 22,814,208 bytes, SHA-256 f5e78fe9ca3c865ab69e57303bebb4db28499dfebf54baf1dc1eb602d381a0db; its published build receipt hash is cb90ed3902942c0770133a4f19ce06077cf66d8e24fa221b96c7bb7e6949000c. Native GUI remains SHA-256 7ece85c6012ae0ae75fc56ed3c3aadb8bc116448ad1782be422ef34a629dc3ce; runtime remains aa282fa47bb2c4e210cea5e470cdbdca58568f0be0bc73e404d86ab9cd95ef6a. Executables and private profile are excluded; receipt paths are original-machine provenance, not portable commands.
archive-manifest.json lists all 204 originals, 3,133,794 bytes (Run 4: 185 files; Run 5: 19). It records exact copies, checkpoints, comparisons, captures and checked published prerequisite Git blobs, plus hashes of this README and raw-evidence attributes. It does not recursively hash itself. No other runs, current working files, binaries, profile or duplicate gate cache are copied. All original bytes were compared again after copying. Seventeen intentional raw .log files, including empty native logs, may need explicit forced Git addition.