fix(native): answer an ancestor chain as a path, not as a set of names

[YevheniiKotyrlo]

Problem

A container is registered by NAME and every ancestor compound is resolved in the subject's own scope, so a chain asks "are both names somewhere above me" rather than Selectors 4 §14.1's narrower question — and .outer .inner .x therefore matches the reversed nesting too.

// src/native/reactivity.ts
export type ContainerContextValue = Record<string, WeakKey>;

// src/native/conditions/container-query.ts
return queries.every((query) => {
  return testContainerQuery(query, inheritedContainers, guards, get);
});

Measured on main with .outer .inner .subject { color: red }:

nesting	should	does
<outer><inner><subject>	match	{"color":"#f00"}
<inner><outer><subject>	not match	{"color":"#f00"}
<inner/><outer><subject>	not match	undefined

The third row is why this is a defect rather than a missing feature: the set is answered correctly and only the ORDER is lost. And it is the shipped group-* path — stacked ancestor variants are what Tailwind emits.

Solution

The scope a container registers in is already in hand at registration time; nothing was keeping it. Carry it, and the chain has something to walk:

export interface ContainerRegistration {
  readonly key: WeakKey;
  readonly scope: ContainerContextValue;
}

testContainerQueries then walks innermost first, resolving each compound in the scope of the one inside it. A single-compound selector resolves exactly as before — every group-hover: / group-data-*: utility in use today.

Worth stating, because the obvious guard shape does not work: a guard recorded against the scope THAT link resolved in makes testGuards — which re-checks container guards against the ELEMENT's scope — disagree every pass, re-rendering until React refuses. Guards stay on the element's own scope, the only one it can re-read and where every chain name is inherited anyway.

Tests

src/__tests__/native/ancestor-chain-order.test.tsx, 13 cases: two pin the compiled shape the walk depends on, eight drive the nesting (ordered, reversed, siblings, intervening elements, a three-deep chain with one pair swapped, both class names on one element), three pin what must not move (a single ancestor, an interaction condition on each compound, an idempotent re-render). The reversed case is the mutation proof — red on main for the reason above.

Every ancestor test in the suite uses ONE ancestor, where a name lookup and a path walk cannot disagree. Nothing rendered two ancestors in both nestings.

Mutation-proved: reverting the src/ diff and re-running these files alone turns 3 of 13 red.

Verification

yarn test 1061 passed · yarn typecheck 0 · yarn lint 0. The three failures are two babel suites that fail identically on an untouched main worktree (Windows-only module-specifier rewrites) — this touches no babel file.

Known limits

Residual, stated: an outer .a shadowed by a nearer .a that is not an ancestor of .b can leave a guard stale — narrower than today's behaviour, which is wrong for every reversed nesting.

Independent of #462, which answers a classless ancestor's condition in the compiler; this answers the ORDER of any chain in the runtime. They touch no common file, so whichever lands last needs no rebase.

No existing issue tracks this — searched the tracker for descendant combinator, container query order, group- and ancestor nesting, zero hits.

Base

Branched off f70c402. main has since taken #451 (a5002c5). 3 of the 5 files this changes also moved there, and 3 genuinely conflict — src/native/conditions/container-query.ts, src/native/react/rules.ts, src/native/reactivity.ts. Every measurement above was taken on f70c402. Say the word and I will re-apply it onto current main.

[YevheniiKotyrlo]

Device evidence — before / after

UNFIXED — The first two bars are red. A chain asks only whether both container names are somewhere above the subject, so .outer .inner .probe matches the reversed nesting too.

FIXED — Only the first bar is red. The chain resolves innermost-first, each compound in the scope of the one inside it — Selectors 4 §14.1's actual question.

before — this build, minus this PR	after — with this PR

Read the ancestor-chain row. One rule governs all three bars — .chain-outer .chain-inner .chain-probe over a blue base — and they differ only in how their ancestors nest: outer > inner, inner > outer, and a sibling subtree holding neither.

The first and third bars are the anti-vacuity pair, and they do not move. The first is red in both frames, so the chain compiled and matched at all; the third is blue in both, so a subject with no such ancestors was never matching. Only the second — the reversed nesting — flips.

That third bar is also why this is a defect rather than a missing feature: the ancestor set is answered correctly today and only the order is lost.

The other five rows are byte-identical between the frames: before is this same build with only this PR's hunk reverted, so exactly one variable differs.

Both frames: pooled Android 36 emulator, 1140×2400 @ 480dpi, dark scheme, same run.