rastrillo / fiesta Public

Test evidence for Task 10 of the Fiesta packaging, copied from the implementer's report so it sits on the branch.

Commits

• 9231830 Port the component drives on their seeded inputs, and hold every example state to the floors
• Fix round 1: e9ba660 Hold the floors drive to its claims, restore the digest's separator, and refresh the weights

Preservation: every ported test passed on its first run

One run, in-tree, before the new tests existed, over the 54 ported tests:

ok  	amadan.net/rastrillo/fiesta/check	89.331s

54 top-level --- PASS, 0 --- FAIL. TestBothSpellingsComputeTheSameStyles logs 1,105 elements at 1280px and 390px, 1,087 differing in the bare pass.

TestTheClassFixtureIsGenuinelyClassSpelled

Its first two in-tree runs were red for reasons in the test (the class scan now considers rst- tokens only; the class attribute comes out of each tag before the scan), then --- PASS (0.01s).

Negative control, rastrillo's mixed classFixture body written over the scratch copy's spelling-class.html:

    spelling_fixture_test.go:83: the class page carries 355 rst- attributes the grammar maps; it is not in the class spelling. The first: rst-box-head in <div rst-box-head>
--- FAIL: TestTheClassFixtureIsGenuinelyClassSpelled (0.01s)

TestEveryExampleStateMeetsTheFloors

First run (passed first time against the tree, as a coverage test of already-floored CSS should):

    coverage_browser_test.go:296: 390 touch: 156 states, 349 controls measured
    coverage_browser_test.go:296: 1280 mouse: 156 states, 357 controls measured
    coverage_browser_test.go:296: 1280 mouse, no scripts: 156 states, 258 controls measured
--- PASS: TestEveryExampleStateMeetsTheFloors (25.92s)

Negative control, --rst-tap: 2.75rem set to 40px in a scratch css/tokens.css:

--- FAIL: TestEveryExampleStateMeetsTheFloors (26.29s)
    --- FAIL: TestEveryExampleStateMeetsTheFloors/390_touch (8.88s)
    --- PASS: TestEveryExampleStateMeetsTheFloors/1280_mouse (8.77s)
    --- PASS: TestEveryExampleStateMeetsTheFloors/1280_mouse,_no_scripts (8.63s)
    coverage_browser_test.go:267: 390 touch, page-header/title-subhead-and-a-primary-action, overlays closed: a[rst-btn=primary] “Write a post” is 147.0×40.0, under the 44×44 floor
    coverage_browser_test.go:267: 390 touch, list-bar/search-with-a-filter-dropdown, overlays closed: input “Search posts” is 257.4×40.0, under the 44×44 floor
(findings in all: 313, naming 92 distinct states)

Preservation control, the same mutation on the ported sizing drive:

--- FAIL: TestEveryTapTargetIsAtLeast44Pixels (5.27s)
    --- FAIL: TestEveryTapTargetIsAtLeast44Pixels/390x844_touch (1.35s)
    --- FAIL: TestEveryTapTargetIsAtLeast44Pixels/1024x768_touch (1.30s)
    --- FAIL: TestEveryTapTargetIsAtLeast44Pixels/600x800_mouse (1.31s)
    --- PASS: TestEveryTapTargetIsAtLeast44Pixels/1280x900_mouse (1.32s)
    sizing_browser_test.go:762: 390x844 touch, overlays closed: input#sizing-bulk-search “Find in the selection” is 200.0×40.0, under the 44×44 floor
(findings in all: 444)

Shards: ~1of3 ok 105.9s, ~2of3 ok 137.1s, ~3of3 ok 126.3s; make ci exit 0, 383s wall.

Fix round 1 (e9ba660)

Important 1: the \x01 separator in digestJS restored as parts.join("\u0001"); a byte-level scan of all 13 ports against their SEED files found that one control byte (SEED markup_v3_browser_test.go line 300) and nothing else. TestBothSpellingsComputeTheSameStyles re-run: --- PASS (3.83s).

Covering run after the rework. First run, with the page-first list lookup (a defect the new check found):

    coverage_browser_test.go:422: 390 touch, field-daterange/date-and-time-seeded: the date field's list (input#dr_ends_at-combo[rst-input][rst-dtp-input] “”) never opened; its options were not measured
--- FAIL: TestEveryExampleStateMeetsTheFloors (65.84s)

Second run, lists scoped to their field:

    coverage_browser_test.go:477: 390 touch: 156 of the manifest's 156 states served, 371 controls measured
    coverage_browser_test.go:477: 1280 mouse: 156 of the manifest's 156 states served, 379 controls measured
    coverage_browser_test.go:477: 1280 mouse, no scripts: 156 of the manifest's 156 states served, 268 controls measured
--- PASS: TestEveryExampleStateMeetsTheFloors (60.26s)

Mutation B (Important 3), the icon-spacing rule's column-gap: 0.3em set to 0 in a scratch tokens.css:

--- FAIL: TestEveryExampleStateMeetsTheFloors (58.73s)
    --- FAIL: TestEveryExampleStateMeetsTheFloors/390_touch (19.86s)
    coverage_browser_test.go:431: 390 touch, list-bar/search-with-a-filter-dropdown, ltr: in a “Paid” the icon is 0.0px from the word before it; it touches the word
    coverage_browser_test.go:431: 390 touch, dropdown/a-filter-menu-with-one-choice-applied, rtl: in a “Newest” the icon is 0.0px from the word before it; it touches the word
touches-the-word findings: 4; other findings: 0

Mutation A (Important 2), first attempt stayed green (<noscript> is ignored because scripting stays on); second attempt reddened the pinned-gaps check instead (the back-nav example links no script). Third attempt, a scratch light.js marking the root and a scratch tokens.css rule html:not([data-rst-js]) [rst-row-action] { display: none !important; }:

--- FAIL: TestEveryExampleStateMeetsTheFloors (58.42s)
    --- PASS: TestEveryExampleStateMeetsTheFloors/390_touch (19.77s)
    --- PASS: TestEveryExampleStateMeetsTheFloors/1280_mouse (19.68s)
    --- FAIL: TestEveryExampleStateMeetsTheFloors/1280_mouse,_no_scripts (18.96s)
    coverage_browser_test.go:449: 1280 mouse, no scripts, list-row-action/lead-marker-status-and-an-action: [rst-row-action] (Row action pill) is measured with scripts and not without; the component does not work with no script

Weights refreshed from a full verbose run (ok 413.4s, 0 failures). Final shard pass: ~1of3 ok 128.4s, ~2of3 ok 145.9s, ~3of3 ok 138.4s; make ci exit 0, 422s wall.