fix(qwik-router): backtrack from static prefixes and rank matches by specificity

[blakeley]

Fixes #8893

Two symptoms

1. A static prefix dead-ends the match. The matcher walks the route trie greedily: at each segment it takes the first child that matches (exact → _M groups → _W → _A) and never reconsiders. So with routes/(marketing)/pricing/index.tsx and routes/(app)/[a]/[b]/[c]/index.tsx, the URL /pricing/x/y matches pricing on the first segment, dead-ends under it, and 404s — even though [a]/[b]/[c] matches the whole URL.

2. No specificity ranking across groups. When two branches both match a URL, the winner is whichever the trie happens to be walked into first, so it depends on group ordering. [x]/static.xml in one group and [a]/[b] in another both match /foo/static.xml, and which one wins is an accident of _M order rather than the more specific route.

Approach

matchRouteTree now resolves the best complete match up front, then replays it:

• findChildAll returns every candidate child for a segment (in the same priority order findChild already walks) rather than just the first.
• findMatchChain does a DFS over those candidates, only accepting a chain that consumes the whole pathname and lands on a node that actually has a route — so a static prefix that dead-ends no longer hides a dynamic sibling.
• compareChains ranks complete chains lexicographically by per-segment kind (static < dynamic < catch-all), so the most specific match wins regardless of group order.

When no chain matches, findMatchChain returns undefined and the segment loop runs exactly as before — the greedy walk, the rest-wildcard fallback, and the 404/boundary recovery are untouched, so unmatched URLs keep producing the same routeParts, params, and error boundaries they do today.

One existing rule is preserved deliberately: an exact route subtree still owns its unmatched descendants, so a dead end under a static route 404s rather than being swallowed by a sibling [...rest] (covered by the existing exact child dead end does not fall back to sibling _M catchall test).

Repro

Minimal reproduction: https://github.com/blakeley/router-matcher-no-backtracking

Tests

Two cases added to packages/qwik-router/src/runtime/src/routing.unit.ts, both of which fail on main and pass with this change:

• a static single-segment route in one group and [a]/[b]/[c] in another — /pricing/x/y reaches the dynamic route instead of 404ing
• [x]/static.xml beats [a]/[b] for /foo/static.xml in both group orders

The rest of the existing matcher suite (the static route itself still winning, plain dynamic matches, catch-all fallback, and an exact dead end not falling through to a sibling catch-all) is unchanged and still green, which is what pins the behaviour this change must not alter.

This restores the documented Qwik 1 routing semantics, where static pages and dynamic route families can share a URL prefix and the most specific route wins.

[changeset-bot]

🦋 Changeset detected

Latest commit: 96c8091

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 5 packages

Name	Type
@qwik.dev/router	Patch
eslint-plugin-qwik	Patch
@qwik.dev/core	Patch
create-qwik	Patch
@qwik.dev/react	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[pkg-pr-new]

Open in StackBlitz

@qwik.dev/core

npm i https://pkg.pr.new/QwikDev/qwik/@qwik.dev/core@8899

@qwik.dev/router

npm i https://pkg.pr.new/QwikDev/qwik/@qwik.dev/router@8899

eslint-plugin-qwik

npm i https://pkg.pr.new/QwikDev/qwik/eslint-plugin-qwik@8899

create-qwik

npm i https://pkg.pr.new/QwikDev/qwik/create-qwik@8899

@qwik.dev/optimizer

npm i https://pkg.pr.new/QwikDev/qwik/@qwik.dev/optimizer@8899

@qwik.dev/devtools

npm i https://pkg.pr.new/QwikDev/qwik/@qwik.dev/devtools@8899

commit: 96c8091