[yeroo]

What and why

The four #1779 cases did not fail because of Windows font metrics or DPI. The GitHub windows-latest runner's desktop is 1024x768, and Windows caps a window at the screen plus its frame. So every call window-size {"w":1400,…} / {"w":1180,…} came out 1028x780: the sweep artifact of run 40195218119 has 268 captures at 1028x780, while the 640/700/460-wide requests are exact. At 1028 px the sheet Formulas tab collapses Formula Auditing, Home collapses Cells/Editing, and the Project Task tab collapses group 4, which are the four failing assertions. The cases are right at the sizes they ask for (Linux passes them).

• ci.yml, ui-sweep-windows: a new step sets the display to 1920x1080 (user32 EnumDisplaySettings/ChangeDisplaySettings through Add-Type, this session only). It reads the mode back and fails the job if it is smaller, printing the modes the adapter offers, so a runner-image change cannot bring the clamp back silently.
• expected-failures-windows.txt: the four Windows UI sweep: ribbon groups clipped or collapsed at CI widths (ribbon-fit, sheet-ribbon, project-plugin-ui) #1779 entries are removed. So are the six the larger screen fixed too, which the first sweep on this PR (run 40209311321) reported as unexpected passes: Windows UI sweep: Project table and Timeline geometry differs (project-selection, project-timeline-box/-grip/-view) #1780 (4 Project table/Timeline cases), Windows UI sweep: page-line-numbers positions differ #1781 (page-line-numbers) and Windows UI sweep: style-gallery shot step errors #1782 (style-gallery). doc-rulers (Leftovers from #755: Leftovers from #699: harness: add save-as, clipboard, fill-handle drag and sheet-ribbon verbs (24 blocked Word/Excel spec cases) #808) still fails as expected, and the Windows UI sweep: border colour asserts are flaky (gantt-scroll, project-cells) #1783 flaky entries stay.
• docs/ui-test-harness.md: states the screen requirement.

Closes #1779
Closes #1780
Closes #1781
Closes #1782

Testing

• python -m unittest discover -s scripts -p "test_*.py": 47 run, OK (10 skipped). YAML parse of ci.yml OK. The step's PowerShell parses, and its C# compiles under pwsh 7.
• This PR's ui sweep (windows) (run 40209311321, head 73c222b): the step logged before: 1024x768, ChangeDisplaySettings: 0, after: 1920x1080, DPI awareness set: True, system DPI 96, logical 1920x1080. Result: 142 scripts, 133 pass, 3 xfail, 0 fail, 0 error, 6 xpass. The four Windows UI sweep: ribbon groups clipped or collapsed at CI widths (ribbon-fit, sheet-ribbon, project-plugin-ui) #1779 cases pass; the 6 xpasses are the entries removed above.

Review

One revmux round (Claude-only profile, 4 reviewers): 0 Major, 1 Minor and 2 Immaterial, all fixed in 73c222b. SetProcessDpiAwarenessContext is now declared bool and its result is printed. The guard also checks the logical screen size at the runner's DPI (≥ 1410x815). The docs and the comment say what is checked.

review stopped: round 1 had no Major; all findings fixed

Follow-up: #1800 (window-size should report the size the window actually got, so a clamp shows in the transcript).

🤖 Generated with Claude Code

https://claude.ai/code/session_01MGUtabxdbfDq8UMHBxPDk3

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[yeroo]

Merged automatically (auto-merge is on for this checkout): review stopped: round 1 had no Major; all findings fixed; merge rounds: none; whole suite green on 00079bb (CI 14/14 checks: Windows UI sweep 142 scripts, 140 pass, 2 xfail, 0 fail, 0 xpass; Linux UI sweep 141 pass, 1 xfail); merge-check ok.

Follow-ups: #1800 (separate issue: window-size should report the size the window actually got). No finding was deferred or disputed.

Also closes #1780, #1781 and #1782. They had the same 1028x780 clamp, and their entries came out as unexpected passes on this PR's first sweep.