/* Shared components. Linked after tokens.css and before any page sheet, so a * page can still override a component but never has to redefine one. * * A selector defined here is defined nowhere else -- that is the rule * tests/test_stylesheet.py enforces, and it is the permanent replacement for * "keep these in sync by hand" comments. */ /* Full-screen modal backdrop. * * Had five definitions. Three were identical (import_instances, rescue_images, * workers/workspace), dirty_harry carried a stripped-down base under its own * .dirty-harry-overlay modifier, and the source-cloud auth partial disagreed: * centered rather than top-anchored, a darker backdrop, and z-index 80. * import.html and workers/workspace.html include both families, so which one * rendered was decided by include order -- undocumented and load-bearing. * * The page variant wins. It is the majority, it is already what operators see * on import and workers, flex-start plus overflow:auto handles a form taller * than the viewport, and z-55 slots into the existing scale (wizard 40, * picker 50, log and form 55, help 58-61). The partial's z-80 wrongly stacked * its modal above the help overlay system. */ .form-overlay { position: fixed; inset: 0; background: rgba(14, 11, 12, 0.38); display: none; align-items: flex-start; justify-content: center; overflow: auto; padding: 40px 24px; z-index: 55; } /* Both activation classes, deliberately. The import and workers scripts toggle * .open; the source-cloud auth partial toggles .active. Honouring only one * would silently break every modal in the other family, which is the most * likely functional break in this phase -- so no classList call anywhere has * to change. */ .form-overlay.open, .form-overlay.active { display: flex; } /* Buttons. * * .btn is worn by both