docs(adr): ADR-0076 OQ#9 裁决 — catch-all 洗白、步骤①吸收完成 (#2462) - #3576
Merged
Conversation
…ep (1) absorbed (#2462) Step (3) decomposed the dispatcher the other way around: 18 domain bodies moved out, dispatch() shrank to gates + a first-match registry, and the if-chain reached zero domain branches. The disease was never the catch-all — it was the unenumerable god if-chain it delegated into. Delegating into a thin gates+registry pipeline is the terminal shape; "kill the dead callable branches" is absorbed (no branches left, only owned registry entries). The adapter-enumerates-registry polish is deliberately not scheduled. Cloud-side verification: objectos-runtime 240 tests (incl. the kernel-resolver strategy) pass unchanged against the decomposed dispatcher. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
|
The latest updates on your projects. Learn more about Vercel for GitHub. 1 Skipped Deployment
|
github-actions
Bot
added
documentation
os-zhuang
added
the
skip-changeset
6 tasks
This was referenced Jul 27, 2026
os-zhuang
added a commit
that referenced
this pull request
Jul 27, 2026
…re splitting into per-prefix mounts (#3608) (#3637) The first place anyone (human or agent) tempted to "fix" the catch-all looks is this line, not the issue tracker. Anchor the #3576 verdict and the #2852 RLS-leak failure mode here; reopen conditions stay archived on #3608 (closing as deliberately-unscheduled). Co-authored-by: Jack Zhuang <277994282+os-zhuang@users.noreply.github.com> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
This was referenced Jul 30, 2026
Closed
os-zhuang
added a commit
that referenced
this pull request
Jul 30, 2026
…oes not own, and unbreak the #4116 ledger (#4117) (#4133) Two things, and the second is urgent. ## The fix (#4117) `app.all(`${prefix}/auth/*`)` claimed a whole namespace and was TERMINAL: it returned the auth service's response unconditionally, including better-auth's 404 for a path it does not implement, and the legacy `handleAuth` bridge's own `handled: false` 404. That is #4088's shape. #4116's scan found it; the manual greps that preceded that scan had not. A 404 from better-auth, or `handled: false` from the dispatcher, now means "not this mount's path" and the handler yields. The predicate is the dispatcher's OWN `handled` flag wherever one exists — an explicit ownership answer beats inferring one from a status; only the better-auth hand-off lacks such a flag, and there the 404 is the signal, exactly as in #4092. What this buys here is narrower than #4092 and the code says so: it does NOT make a later-registered Hono route reachable, because the `${prefix}/*` dispatcher catch-all below is deliberately terminal (ADR-0076 OQ#9, #3576/#3608) and would swallow one either way. It means an unowned `/auth/*` path now reaches that gated `dispatch()` instead of dead-ending in a 404 built one mount earlier, so a domain handler registered for it becomes reachable — which is this adapter's actual extension mechanism. The first draft of the tests asserted the #4092 outcome and failed, which is how the distinction got established rather than assumed. ## Unbreaking main #4122 shipped the ledger declaring TWO ratcheted mounts here. Between its base (974c6d4) and its merge, #4087/#4112 landed and DELETED the `/storage/*` bridge — so the squash produced a main where the ledger declares a mount that no longer exists, and `check:wildcard-fallthrough` fails on main with "DECLARED but not found". That is my merge race, and this commit removes the entry. Worth recording why the guard behaved well here: the stale-entry arm is what caught it, immediately, on the first run against the new main. And #4112 reached the same verdict about that wildcard independently, from the other direction — "the wildcard was wider than the two routes it served". Two independent reads landing on one defect is the argument for enumerating the shape rather than finding it by eye each time. Also extends `callsContinuation`: handing the continuation to a helper (`yieldUnowned(c, next, …)`) now counts as yielding. Without that the checker called this fix terminal — a false negative that would have pushed the compose subtlety toward being duplicated at each call site instead of written once. Guard: 7 yielding / 0 ratcheted / 5 exempt over 12 mounts. Self-test 17 cases. adapters/hono 73 passed. Verified both defences bite: reverting the fix fails the guard with the TERMINAL diagnostic and fails 2 of the 6 new tests. The 3 `tsc` errors and 2 `eslint` rule-definition errors in this package are pre-existing, measured identical on a clean tree. Claude-Session: https://claude.ai/code/session_013VZLsKypjrGhiLpio2uWKu Co-authored-by: Claude <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
裁决（防止将来执行者按旧文本行动）
ADR D11/OQ#9 原文写"until the catch-all is retired by step ③"——步骤③ 用相反的刀法达成了目标：18 个域体全部搬出、dispatch() 缩成关卡序 + first-match 注册表、if 链域分支清零。
病根从来不是 catch-all 本身，而是它委托进的那个不可枚举的上帝 if 链（"declared≠mounted≠implemented" 的 v16 缝 bug 类就藏在那里）。委托进"关卡+注册表"薄管线是合理终态：路由表可枚举（
DomainHandlerRegistry.list()）、每条有 owner 模块与接缝测试、cloud fallback 织物（/share-links per-env 主面、/notifications 唯一 owner）走同一管线并被 conformance 41 断言 + dogfood 351 端到端锁定。步骤①（"删死分支"）被吸收：已无分支可删，只有有主的注册表条目。cloud 侧验证：objectos-runtime 240 测试（含 kernel-resolver 策略）对拆解后 dispatcher 原样通过。
可选 polish（adapter 枚举注册表挂显式路由）需把关卡序 middleware 化——成本大于收益，有意不排期（conformance + route-parity 门在位）。
纯 docs（skip-changeset）。关联 #2462。
🤖 Generated with Claude Code