4.3 KiB
Process removal compact review follow-up
The unchanged Windows build from 572e96ba03522cfb822b13fe4abb50e857a5de48
supports the bounded OA derived Process removal workflow at 1102x732.
This closes the compact review gap in the original report.
The contribution remains local-only pending recovery of Gitea writes; no
integration, issue closure or release is claimed.
Observed behavior
Run 3 opened an isolated copy of the original fixture. Fit made its four cards and authored link visible. Context Add formed the exact group Action @a, Control @c and SequenceLink @approved, with the link primary. The review's title, aggregate counts, search, paging and fixed Cancel/Remove buttons fit within the window. Scrolling page 1 exposes the explicit-before/implicit-after order changes without losing the footer. Page 2 shows both remaining exchange positions and full direct/transitive caller names and aliases. Search finds @outer; a nonmatching query displays recovery guidance.
Cancel preserves all three selected members and the link primary. Save gives the complete 2609-byte baseline. A fresh review resets search; explicit Remove leaves two exchange cards with their implicit connection and clears selection. Save equals the independently specified 2258-byte v2 result. One Ctrl+Z and Save restores every original byte. The normal close exits 0, with empty logs and all twelve immutable file pins unchanged. Redo and fresh reopen retain their original runs 1–2 evidence; they were not rerun here.
- Baseline SHA256:
9965f4024584d64ac81442724aeb77b018b0cfae94e50bd412c5dcbbac49f6bf. - Removed SHA256:
b4a2b07ad4c796f555c61945ca42b1c221b454b35488c60fefc86f96f7aa71ba.
See the parent action record and independent review. Actions reported by the parent are distinguished from screenshots and source receipts. The setup record preserves ineffective edge drags, the unsupported UIA value operation and picker-focus reporting inconsistency. None is counted as successful application interaction.
Evidence and limitations
The new native evidence snapshot
includes all three closed runs and the original failed gate history. It does
not overwrite the earlier two-run package. Its manifest
pins every included member and identifies omitted executable/DLL bytes.
The standalone verifier passes: 461 members, eight omitted binary payloads,
three accepted closed native runs, and the retained preflight/gate history.
ZIP size is 9,190,152 bytes; SHA256
02ef32920cb76e12da8b2e41cdecd0d5eea5526371df6425462195c30b564250.
The independent archive audit
compares all members with original records, rehashes omitted local binaries,
confirms preservation of the original two runs, and separately runs the verifier.
Recheck with Python 3.11+:
python -B harness/process-group-removal-evidence-20260925/compact-native/verify.py harness/process-group-removal-evidence-20260925/compact-native
This is compact removal review acceptance for one OA derived fixture, not complete compact authoring or accessibility acceptance. At 100% zoom the narrow canvas crops the graph; Fit reduces label size. The narrow tool palette uses abbreviated labels. Native Shift-click, keyboard-only navigation, equivalent SA native acceptance, full-suite and Linux checks remain unclaimed. No production source changed and no new Cargo build was needed for this native trial.
A separate static investigation found no legitimate growing-source Process removal for a genuine size-budget failure. Valid removals prune admitted contributions, links and references; Scenario's optional-reference expansion fixture does not transfer to Process. The existing Process test exercises an injected engine failure. The genuine Scenario test covers the shared Document source-budget boundary. Neither is relabeled as actual Process source-write-failure coverage, and no artificial fixture or production fault seam was introduced.