[modacker]

Change

The macOS and Windows pixel gates fail on 15 shots each after PR #36, because that pull request re-recorded the Linux baselines only and the darwin and win32 baselines still hold the picture from before the UI changed.

This pull request re-records the 30 stale baselines: the four plugin catalogue forms, the shell and the workspace panel at 1440, 1280, 1024, 900 and 768, and the shell rail, the transcript and the progress panel, for darwin and for win32.

The images are the actual images of run 42998849283, the run that failed. A baseline recorded on any other machine is that machine's rasteriser's idea of the same picture, and the runner that compares it would fail it, so the images have to come from the runner itself.

Validation

• Checks run and results: the browser suite of run 42998849283 failed on exactly these 15 shots on macOS and on Windows. The failing shots are named in the job log and the artifacts carry their actual, diff and error-context files.
• Each of the 30 new baselines was checked against the image it replaces: every darwin and win32 shot has the same pixel dimensions as the Linux baseline of the same shot, all 15 of them on both platforms. The progress panel is 288 by 139 pixels now and was 288 by 281, because PR fix(webui): repair file opening, panel interactions and CI diagnostics #36 sized the progress sections to their content instead of holding a fixed 200 pixels.
• Every other shot moves by the same few pixels: 1578 on the four dialogues and 2126 on the eleven shells, rails and transcripts on darwin, 1644 and 2204 on win32. That is 0.12% to 0.32% of the picture, and the count is the same across the shots because the same rail rows and the same lines of dialogue text are what moved.
• The platforms agree with each other better than they did before. The gap between the darwin and the linux picture was 0.31% to 0.68% of the image and is 0.20% to 0.47% now; on win32 it was 0.30% to 0.68% and is 0.21% to 0.48%. What is left is anti-aliasing rather than a different layout.
• Reproduce the numbers with magick compare -metric AE over the old and the new file of each shot; the ratio is printed in parentheses next to the pixel count.
• Performance: basic. No perf:full run: this pull request replaces test images and touches no product code.
• NOT RUN, platform limitations and live-service boundaries: the shots can only be judged on the three runners. A baseline recorded on any other machine, including a developer laptop, is that machine's rendering and would fail the runner that compares it.

Publication and contribution checks

• I have permission to contribute these changes under the existing licenses applicable to the changed files/packages; imported material and its provenance are identified and existing notices are preserved.
• No credentials, account data, real user content, internal source history or private review material is included.
• Added/removed source files were reviewed before regenerating release/public-source.json; new tests are declared in test/vitest-suites.json where applicable. No source file and no test file was added or removed: only 30 baseline images under packages/webui/test/browser/ were replaced, so neither registry changes.
• Shared English/Chinese documentation and capability/verification records are updated where applicable. Mock/offline results are not described as live-service acceptance. No document describes these baselines, so none changed.

Maintainer handoff

Publication scope or license changes (if any): none.

Shared-source port: not needed. Record public PR references only; keep private links and review material out of this PR.