Skip to content

Latest commit

 

History

History
49 lines (41 loc) · 4.81 KB

MACOS_ACCEPTANCE.md

File metadata and controls

49 lines (41 loc) · 4.81 KB

Native evidence reconciliation and delta acceptance

Updated 2026-10-14 for local Phase 1, based on private development main 5d81d3f plus the scoped export/packaging/documentation delta. This worksheet records reused receipts and new checks; it does not request a blanket acceptance rerun.

The revised handoff pin aa2e75c5 has matching affected ReportDestination source, existing destination tests and package_source script except trailing blank lines. The live checkout already contains Reload/Retry and Input Identity. The historical consultant v0.1.0 patch stack was not applied. Optional Current-only Command-R and accessibility changes are excluded. Info.plist remains v0.1.2 build 3.

Evidence and decisions

Capability Accepted receipt / source identity Phase 1 material delta Decision and boundary
Baseline native/core integration evidence/32, original 58 methods in debug and release including eight native methods; Apple silicon, macOS 27.2, Apple Swift 6.4 Export ancestry policy only Reuse unchanged coverage; no full suite
Export boundary historical evidence/11–12, 23–26; v0.1.0 quarantined native export 81–83 Retain entry ancestry as well as resolved target New synthetic policy red/green 122–123; all four affected destination methods pass on macOS; not an exploit or concurrent-write reproduction
Current application build metadata development build 105; release 111 Core export source changed New real development build/launch 126, macOS 27.2 build 26B5112a, Apple Swift 6.4.2, arm64; ad-hoc staging is not distribution signing
Ordinary save-panel export accepted native export 81–83 Production guard now checks both path interpretations New synthetic-comparison save-panel export 129; new file created. Computer Use timed out reading completion status. Process sample 130 found the main thread waiting in the normal AppKit event loop; this does not establish all UI responsiveness. No new live-input denial/failure-recovery claim; production source retains validation before encoding/write and error/busy cleanup
Drops/replacement/cancellation 44–52, browser 41–43, inspector 54–63 None Reuse accepted receipts; no gesture/corpus rerun
Reload/Retry e0f5011, focused 87 and native 89–93 None Reuse; no optional command change
Input Identity 51f058e, six focused methods 104 and native 105–110 None Reuse; UI identity stays out of schema 1
Report contents/privacy earlier report checks and exact unchanged bytes/schema test 104 No schema or encoder change Reuse full-report/path-free invariants
Source packaging supplied candidate seven manifest methods, Linux Python 3.12, copied receipt 124 Adoption of index/required-doc/versioned-output guard; docs copied verbatim Reuse five unchanged methods; two temp fixtures resolve macOS roots and pass 128; new cleanup/refusal checks 125/127
v0.1.2 distribution/install sanitized tag 9b91dd7, evidence 111–121; local release record 5d81d3f None to existing assets Reuse signing/notary/ticket/Gatekeeper/download/install receipts; no rehash, re-signing or repackaging
Window lifecycle and commands baseline native acceptance and inspector/reload receipts None Reuse; ordinary launch only accompanies changed core build
Minimum OS / Intel hosts target macOS 14+, arm64 distribution; x86_64 input inspection is separate None Minimum-OS runtime, Intel-host and clean-Mac acceptance remain unestablished
Accessibility / alternate appearance Existing features retained Excluded Work remains stopped; no complete acceptance claim

New-check details and limitations

Only changed export policy and packager outputs justify new checks. The regression uses synthetic entry and target URLs, with no link/write fixture. It first failed as an assertion with the target-only policy, then passed after both ancestry checks were retained. Component identity and device/inode protection remain present. The guard is a preflight, not race-proof containment or artifact custody.

New packaging tests use mocked index metadata and temporary synthetic files. Initial cleanup tests stopped at preflight because macOS temporary roots resolve through /var; the test roots were corrected, then only failed cases reran (127). The two supplied temporary-file tests were similarly adapted and passed once on macOS (128). The other five accepted manifest cases were not repeated. No real source ZIP was regenerated. No published app, installed app, signature or old receipt was changed.

Raw evidence stays local and is excluded from source packaging. Public validation summaries describe accepted observations without claiming a fresh complete audit. A future release requires separate authorization and first checks for its new artifact; this worksheet itself authorizes no push, signing or publication.