[michaldengusiak]

Owner acceptance and authorized merge - 2026-10-18

The owner has accepted the documented explicit-EPW numerical residuals for G1 integration and authorized merging #237 followed by #238 after passing checks. This supersedes all earlier HOLD wording below. No GitHub approval review is fabricated and no repository protection is bypassed.

This is a one-time acceptance of the prescribed two-cell/Boston evidence, not a change to comparator tolerances or a claim of bit identity. Maximum energy residual is 3.134337021037936e-7 kWh, maximum peak residual 2.735980331181054e-10 kW; maximum relative residual about 1.9433e-11. CLI-only repeats also vary and all 163 normalized IDF objects have matching engineering fields. The precise source of run-to-run variation is not claimed proven.

Real native cancellation/process cleanup, concurrent-edit isolation, the 608/608 exact persisted Tas baseline comparison and the actually executed real-run test are PASS. Evidence and limitations are committed in documentation/OpenStudio-Simulation-PR229.md and documentation/evidence/g1-20261018/. No new simulation or local broad suite was run for this acceptance-only update; earlier local/full-suite/engine results remain attributed to their tested content.

Current documentation-only final candidate: 9abfc03469bb46c29e46b2dd3feb1961e09f8a5b. Final-head CI pending. #237 merges first; #238 stays draft until dependency and final gates pass. Numerical engine disagreement, HungaryHouse Tas crash, missing input manifests and producer deployment remain separate follow-ups.

---

Earlier implementation and acceptance history follows.

Closes #229 (part of #227; tracker SAM-BIM/SAM#206).

Behavior

Adds Simulate > OpenStudio Simulation beside Energy Simulation. The run freezes a deep copy, uses SAM_OpenStudio's existing conversion/results pipeline, and writes a separate result model, manifest, weather, OSM/OSW and SQL. The open model and existing Tas workflow are preserved. #238 adds the unified selector and comparison workflow after this PR.

Current acceptance: HOLD - 2026-10-18

Owner authorization covers focused testing, evidence-only final heads and later merging after all gates pass. No additional pin approval is needed solely for these evidence commits. The required explicit-EPW equality gate remains unmet; no acceptance waiver or numerical tolerance has been introduced.

Final evidence-only head: b4c5ad9279c58b0a5db0874333db2a696106b145. Freshly tested UI source: b7483443f53dd53a4d16f159d6bc327c901a7770; same-dependency pre-PR Q4 baseline: ebe5e09b30b95e43e5d047b3aef2d8c5f01110e9. Both Release builds succeeded. Final diff is documentation/evidence only; git diff b7483443 b4c5ad92 --check passes. No product fix was made.

Newly executed check	Result
Real native cancellation	PASS: Cancel while run-owned OpenStudio/EnergyPlus processes were observed alive; both terminate, manifest Cancelled, exit -1, no completed result model. No external kill used.
Concurrent-edit isolation	PASS: rename committed before run completion; snapshot/result retain old space name, saved disposable live model retains edit; snapshot hash unchanged.
Existing Energy Simulation vs pre-PR Q4	PASS for the box case: same model/weather/settings/dependency binaries; 608/608 persisted numerical values exactly equal. Does not claim byte-identical TSDs or full hourly-native audit.
Prescribed two-cell + explicit Boston EPW	Both runs succeed; UI records Weather=File, exit 0, 13 warnings, 0 severe/fatal, 31 results. Source/weather provenance matches. Four model energy/peak values are not identical: HOLD under original G1 step 12.
Opt-in RealRun test	PASS 1/1, actually executed with valid model/EPW/CLI; detailed output Succeeded, no NOT RUN, TRX verified.

Standalone/benchmark stored values and deltas:

Metric	Standalone	Benchmark	Difference
Heating kWh	31072.86245449682	31072.862454810253	-3.134337021037936e-7
Cooling kWh	5764.530483239447	5764.530483144583	9.486393537372351e-8
Peak heating kW	19.82144246485972	19.821442464872938	-1.3216094885137863e-11
Peak cooling kW	14.079292685507252	14.07929268578085	-2.735980331181054e-10

A second CLI-only run also varies (heating delta -6.598711479455233e-7 kWh versus CLI 1). All three generated IDFs have identical engineering fields after matching generated aperture/frame names and object order: 163 objects, normalized SHA-256 3b0dbe8f96a18c8b99b311811b2cccdc15c7b34046a549a0981a385cdf0da397, without rounding or dropping physical fields. This supports existing same-engine conversion/order reproducibility rather than a newly established UI mapping defect, but does not meet literal identity. No physics, units, tolerances, comparison arithmetic or deployment changed.

Exact pins, commands, input hashes, assertions, screenshots and limitations: PR record, section Windows G1 acceptance execution; portable evidence. Full machine-local runs/builds/logs/TRX: C:\TasOut\pr237-g1-20261018 (must be copied/provisioned on another machine). Original fixtures preserved; only disposable HungaryHouse copy renamed.

All 14 dependency source pins were held fixed on sow/2026-Q4; eight engine/core/interop DLL hashes match across the two UI builds. Important pins: SAM 618db819300df8b698b3bbe5174a203244134738, SAM_OpenStudio d4216f3525c01ca762c1b3e9b6d099733d037f54, SAM_Tas a34a298dde2803de514081a0ed72ac3ff790997b, SAM_Validation 68eded0060047af714853a02a152fb9c960bfeb1, containing #15 merge 912ae94ef68af20b437fc92ec7e8891fd1cb640f. Runtime OpenStudio 3.10 / EnergyPlus 25.1 / Tas 9.5.7 / Windows x64. SAM_COMMIT=618db819300df8b698b3bbe5174a203244134738 supplied for benchmark provenance.

Historical successful Both/partial, saved box equality, 2835/2835 suite and presentation evidence were inspected, not rerun. These do not waive the failed explicit-EPW requirement. The final-head hosted workflow builds but does not run WPF tests.

Final-head CI and merge gates

Build 39113785793 PASS; SPDX 39113785833 PASS, at final head b4c5ad9279c58b0a5db0874333db2a696106b145. Completed build log inspected: all 14 dependencies select sow/2026-Q4, including SAM_Validation, SAM_SQLite, SAM_OpenStudio and SAM_Tas. No Validation feature branch is selected. Log preserved at C:\TasOut\pr237-g1-20261018\logs\pr237-final-build-39113785793.log.

Live metadata: OPEN, non-draft, no submitted reviews or review threads; base sow/2026-Q4. Owner merge authorization does not bypass acceptance/protection. #238 is held dependent on this PR. No UI merge, PROJECT_PROGRESS closeout or task-branch cleanup has occurred; checkouts clean, evidence committed and pushed.

Next: owner/domain review of the strict identical-value specification, or a separately reviewed deterministic-conversion fix if bit identity is required. Do not merge either UI PR until that acceptance decision is resolved. Numerical Tas/OpenStudio disagreement, HungaryHouse Tas crash, unavailable input manifests and producer deployment remain separate follow-ups.

---

The original G1 procedure below is retained as the acceptance source; its step-12 identity requirement has not been relaxed. Current results are above and in the dated committed record.

Manual G1 acceptance (owner, Windows x64)

Prerequisites

1. OpenStudio 3.10.0 installed. In a new terminal, openstudio --version should print 3.10.0. If it is not on PATH, the default install C:\openstudio-3.10.0\bin is also found.

2. Build this branch with the usual sibling layout. Either:

   • SAM_Deploy BuildAll, which builds SAM_OpenStudio with Platform=x64 before SAM_UI; or
   • by hand, in this order: SAM, SAM_SQLite, SAM_OpenStudio with /p:Platform=x64, the other dependencies, then SAM_UI (claude/openstudio-mcp-server-1jmjlo).

   The SAM_UI build fails early with a clear message if the native OpenStudio DLLs are missing.

3. Check the deployed folder. After the build, %APPDATA%\SAM (or wherever you launch SAM Analytical.exe from) must contain:

   • openstudiolib.dll, openstudio_csharp.dll, openstudio_model_csharp.dll, openstudio_translators_csharp.dll;
   • OpenStudio.dll, SAM.Analytical.OpenStudio.dll, System.Data.SQLite.dll;
   • runtimes\win-x64\native\SQLite.Interop.dll.

A. Real run with embedded weather
4. Copy SAM_OpenStudio\tests\resources\models\rawModel-WindowsWeatherDataDDYAdiabaticAirPanel.sam to a scratch folder, open it in SAM Analytical, and Save it. Note the file's SHA-256 with certutil -hashfile <file> SHA256.
5. Run Simulate > OpenStudio Simulation. Pick the scratch folder, then answer Yes (use the model's weather data). The progress window should step through: Freezing/converting, Saving OSM, Writing OSW, Running OpenStudio/EnergyPlus, Reading results. Check that the main window stays responsive while it runs.
6. Expected summary: finished, a run folder <model>_OpenStudio_<timestamp>_<id>, a result model with N results, Weather: Embedded, and EnergyPlus: exit code 0, … 0 severe, 0 fatal.
7. In the run folder, check that these exist:

• snapshot.json, weather.epw, <model>.osm, <model>.osw;
• run\eplusout.sql, run\eplusout.err;
• <model>_OpenStudio.json, openstudio-run.json.

In openstudio-run.json, status should be Succeeded, engine.fatalCount should be 0, and engine.cliPath should be your 3.10 openstudio.exe. Report severeCount and warningCount.
8. Original model preserved:

• The title bar or undo history shows no new modification.
• Save again and re-hash: the hash equals step 4's. (If SAM writes a timestamp on save, compare in a JSON diff instead; the only expected difference is the timestamp.)
• Energy Simulation results already in the model (if any) are still there.

9. Result model correct:
   • Open <model>_OpenStudio.json in SAM Analytical. Space results show for Cell 1 and Cell 2.
   • Results > Remove lists only … - SAM.Analytical.OpenStudio entries, with no Tas results.
   • The model has one Analytical Model Simulation Result (heating/cooling kWh and peak kW).

B. Explicit EPW, compared with the benchmark CLI (same code path)
10. Repeat step 5 but answer No and pick SAM_OpenStudio\tests\resources\weather\USA_MA_Boston-Logan.Intl.AP.725090_TMYx.2004-2018.epw. The summary should say Weather: File.
11. Run benchmark-openstudio --model <fixture>.sam --weather <same epw> --out bench.json. The exe is under SAM_OpenStudio\benchmark\SAM.Analytical.OpenStudio.Benchmark\bin\.
12. Compare bench.json model.consumptionHeating/consumptionCooling/peakHeatingLoad/peakCoolingLoad with the result model's Analytical Model Simulation Result. They should be identical: same pipeline, same inputs.

C. Cancel and isolation
13. Start another run, then press Cancel while it shows Running OpenStudio/EnergyPlus. Expected:

• openstudio.exe and energyplus.exe disappear in Task Manager;
• the summary says cancelled;
• the manifest status is Cancelled;
• there is no <model>_OpenStudio.json.

14. Start a run and, while it runs, edit the open model, for example rename a space. The run completes. Its snapshot.json and result model have the old name, and the open model keeps your edit.

D. Tas unchanged
15. On the same model, run Simulate > Energy Simulation (Tas) with your usual settings, once with this build and once with a sow/2026-Q4 build. The dialog and workflow should be identical, and the results equal within Tas run-to-run reproducibility.

E. Failure reporting
16. (Optional) Temporarily rename the OpenStudio install so openstudio.exe is not found, then run. Expected: the summary says failed with "openstudio.exe was not found …" and SAM_UI does not crash. Restore the install afterwards.
17. Run dotnet test WPF/SAM.Analytical.UI.WPF.Tests -c Release. The full suite should pass. OpenStudioSimulationTests.RealRun_… must not print NOT RUN.