[ellenbauer]

Refs #2726. Follow-up to #2742.

What was wrong

#2742 records core's flow layout for a list navigation whose items all declare a block-level display. On the real Dunbar Oaks (Wix) phone menu it does not fire, so "Annual Meetings" and "News & Events" are still cut off.

The check read each item's display with resolvedConditionalDisplay(). That takes the last matching declaration in source order and ignores specificity. The menu items match two device-scoped rules:

• :where([data-dla-device-document="mobile"]) .BZjmPL>ul li{display:block}, specificity (0,1,2), earlier;
• :where([data-dla-device-document="mobile"]) .FWN1UT{...;display:flex}, specificity (0,1,0), later.

The browser renders the items as block. The resolver returned flex, so the check saw flex items and core/navigation kept its flex row. The #2742 test missed this because its fixture had no competing weaker item rule.

What changed

• NavigationPatternContext::cascadeDisplay(): the display from the specificity-ranked cascade over static and applying conditional rules (StyleResolver::controlSurfaceResolvedStyle()). It is a recognition signal only and is never projected, which is what that resolver is documented for.
• NavigationPattern::withStackedBlockListLayout() reads the host, list and item displays through it.

Tests

• tests/unit/navigation-stacked-block-list-layout.php now adds the Wix shape, a later and weaker .item{display:flex}, and a flex-items case: 8 assertions. On trunk 4 fail; all pass here.
• capture-evidence-equivalence: the same nine cases as fix(php-transformer): stack declared block menu items in core flow layout #2742 move their style-rule candidate cache counters again. Replaying them on trunk and on this branch and diffing the normalized output shows no other change. Re-recorded and noted in the README.
• composer test (memory_limit -1) passes on this branch, rebased on trunk 2a963f0 (v0.40.0).

Evidence on the real page

The Dunbar capture has since been removed from the scratch area, and new captures are on hold while the capture vocabulary changes (#2725), so I could not re-import Dunbar with this branch. What I have:

• Instrumenting the check on the real Dunbar home page (header part inlined) before this change printed items=li:flex x6 (host and list display empty). With cascadeDisplay() all six resolve to block and the layout is recorded.
• Studio import with everything merged so far (fix(php-transformer): stack declared block menu items in core flow layout #2742, fix(php-transformer): keep wrapped peer menu items flat and off panel classes #2743, fix(php-transformer): keep the source panel box on a projected modal dialog #2744) on the same capture, iPhone user agent at 375 (localhost:8940): the dialog is now at x=27.5, 320 wide, no padding or border, beige layer rgb(233, 220, 211), but the nav is still display:flex and the items are 100px wide at x=165, with "Annual Meetings" and "News & Events" clipped.
• Same page with the nav switched to core's flow layout classes in the browser, which is what this change emits: items 155px at x=110, all six labels full, matching the source capture.
• 1280 and 375 closed: 0 differing pixels between the trunk import (localhost:8914) and the merged import (localhost:8940). The menu is only in the phone dialog, so this change does not reach either view.

Regressions

• Only list navigations whose cascade gives every item a block-level display with no display on the nav or its list change. The capture-evidence corpus (349 cases) shows no output change beyond cache counters.
• I could not run fresh Studio imports for other sites: the captures here use the old capture vocabulary that trunk no longer reads (refactor: neutral capture vocabulary for Blocks Engine capture evidence #2725), and the current Studio CLI rejects the older importer packages.

Import proof (workbench, new-vocabulary captures)

Studio CLI create --from with the liberation workbench captures (DLA main 569c336). The baseline is SSI v1.31.2 with BE trunk 2a963f0 (wb-*-trunk). The proof package is the same ZIP with #2780, #2781 and #2783 merged on 2a963f0 (ssi-be-proof-2726-combined-e51205f.zip). iPhone user agent at 375; desktop at 1280. Quality gate passed on every import (quality_pass: true, fail_import: false). The proof sites are stopped, not deleted.

• Dunbar Oaks, trunk (localhost:8952) against proof (localhost:8953), phone menu open:
  • trunk: the nav is display:flex, items are 100px wide at x=165, and "Annual Meetings" and "News & Events" are clipped;
  • proof: the nav is display:block, items are 155px wide at x=110, and all six labels show in full, matching the source.
• Dunbar at 1280 and at 375 with the menu closed: 0 differing pixels.
• stevenheaney90, trunk (localhost:8956) against proof (localhost:8957): the same Wix menu nav is now flow instead of flex. It is not visible on that page (no captured dialog), so 1280 shows 0 differing pixels. At 375, one of two screenshot runs is identical; the other differs only in a time-varying block (x 20 to 146, y 1300 to 1422) that also differs between two runs of the same site.

🤖 Generated with Claude Code

[ellenbauer]

Merged after an independent review (approve). Follow-up to #2742: host, list and item displays are read from the specificity-ranked cascade as a recognition signal only, so Dunbar Oaks' phone menu gets flow layout and all six labels show in full. Import-proven on a fresh new-vocabulary capture: labels 155px wide, 1280 and 375-closed pixel-identical to trunk. Refs #2726. cc @chubes4