[michaldengusiak]

Roadmap S05 - Licensed Phase A GUI acceptance (SAM documentation/SIMULATION_ROADMAP_Q4_2026.md). Record and evidence only: no product code changes. Record: documentation/S05-LicensedAcceptance.md. Evidence: documentation/evidence/s05-20261020/.

Base sow/2026-Q4 ad1f2486, which has S01-S04 merged (S03 #242 1e460bbe, S04 #243 307cb0e3). The preflight (#244 46a43d8d) was a rehearsal and is kept separate.

Build under test

• Fresh build. The dependency stack was rebuilt at the Q4 tips: BuildAlls_v5 skip2027, 26 repositories, exit 0. The app was then built as Release into an isolated folder.
• Isolation. APPDATA and USERPROFILE were redirected to a sandbox. Neither the installed SAM nor the real %APPDATA%\SAM was touched.
• Finding: the old local build was pre-F1. Before this rebuild, the laptop's SAM\build predated SAM#212 (F1). All overnight real-engine evidence had therefore used a pre-F1 SAM.Weather.dll. In this build, SAM.Weather, SAM.Analytical and SAM.Core are byte-identical across the app, both producers and SAM\build.
• SAM_COMMIT. Per the owner decision of 2026-10-20 (temporary, until S06), it was set only for Both runs, to SAM 62b54156. That is the clean commit the producers were verified to be built from. The manifests record it as a manual override.

Matrix: all nine cases PASS (real ribbon, licensed TAS 9.5.7, OpenStudio 3.10 / E+ 25.1)

#	Case	Result
1	Tas only	PASS. Succeeded, full provenance. Persisted results 602/602 exact vs the S01b baseline.
2	OpenStudio only, explicit EPW	PASS. Succeeded; CLI hash, E+ version from eplusout.err. Snapshot without view state; result model with it.
3	Both, explicit EPW, two sessions	PASS. Compared in both. Snapshot 0c30db14… and canonical 05c1e334… identical. comparison.csv identical to S01b B2 and the preflight.
4	Both, embedded weather (post-F1)	PASS. The exported EPW is byte-identical to SAM#212's embedded-corrected.epw; the preflight's matched embedded-before.epw.
5	Both, explicit EPW	PASS (case 3 runs). Weather File 472ffb4e… in the manifest and in both producers.
6	Multizone (SAM_OpenStudio's committed two-cell fixture), two sessions	PASS. Compared; 2 spaces matched; identical snapshot and canonical hash. E+ values vary by about 1e-11 relative between sessions; this is known for this model.
7	Injected Tas refusal	PASS. One "Simulation failed" box (return code 0, error log, TAS-SIM-001). tas-run.json Failed with the same text.
8	Cancel (Both during Tas; OpenStudio during EnergyPlus)	PASS. Both runs Cancelled; no comparison and no result model; no engine process left.
9	Concurrent start	PASS. Energy Simulation, Run Simulation, OpenStudio Simulation and the comparison window's Run were each refused with the S02 message. Only the first runs' folders exist.

Test-harness defects (lost folder-dialog click, sandbox Desktop missing, BM_CLICK on an inactive box, a miscounted process wait, comparison-window refusal looked for at the wrong depth) are classified in the record. Each was fixed in the harness and the affected case rerun; attempts are kept as evidence. No product defect found.

Not exercised in the GUI: the Optimisation "Run Energy Simulation…" follow-up (the same guarded SimulateTas path, unit-tested in S02). Part O Hub is unguarded (S02 owner decision).

Final PR head 4f0d12b0: hosted Build (39378742481) PASS, SPDX (39378742474) PASS. 219f442b is the record and evidence; 4f0d12b0 adds the local unit-suite result on the rebuilt dependencies: 2926/2926.

🤖 Generated with Claude Code