Skip to content

Commit 89cf070

Browse files
committed
Merge remote-tracking branch 'origin/main' into claude/project-thread-nye4b2
Claude-Session: https://claude.ai/code/session_01FEkQ5nSMgkteiemkoN31vf Co-authored-by: Claude <noreply@anthropic.com>
2 parents e1583eb + 6cbdef1 commit 89cf070

57 files changed

Lines changed: 3300 additions & 158 deletions

File tree

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.
Lines changed: 19 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,19 @@
1+
---
2+
'@objectstack/formula': patch
3+
---
4+
5+
feat(formula): a CEL syntax refusal keeps the engine's own fault code and the source range it points at, so an editor can say what is wrong and where in its own words (#27904)
6+
7+
Clause-②: no
8+
Metadata-protocol: none
9+
10+
`validateExpression`'s catch-all CEL refusal, `invalid-cel`, now carries two optional params beside `role` and `detail`:
11+
12+
- `celCode` — cel-js's own code for the fault, read off its error object: the parser's for a syntax fault (`unexpected_token` when an operand or the end is missing, `expected_token` when a closing `)` / `]` / `}` is missing, `unexpected_character`, `unterminated_string`, …) or the checker's for a type fault (`no_such_overload`, …).
13+
- `pos` — `{ start, end }`, the UTF-16 offsets into `source` that the engine points at (`end` exclusive). A missing `)` reads `start === end === source.length`.
14+
15+
Both are absent when the engine reported none, and `pos` is absent when the engine read a source the platform rewrote first (the null-guard ternary lowering), since a range into that text would point at characters the author never wrote. A guessed `0` is never written.
16+
17+
`EvalError.pos`, declared before and never filled, is now filled by `celEngine.compile` and `celEngine.evaluate` for a cel-js parse, check or evaluation fault that carries a range, on the same terms.
18+
19+
The refusal's `message` is unchanged, byte for byte, and so are the codes and the params of every other refusal.
Lines changed: 17 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,17 @@
1+
---
2+
'@object-ui/app-shell': patch
3+
'@object-ui/i18n': patch
4+
---
5+
6+
feat(studio): "New similar" on a read-only base's permission set in the Access pillar, opening the copy where the clone door landed it
7+
8+
Clause-②: no
9+
Metadata-protocol: none
10+
11+
A customer's own permission set is a new set in the package that holds their changes (ADR-0146 D6). Since the clone door lands a copy of a read-only package X's set in X's paired `X.custom`, Studio can offer that gesture where the base is designed.
12+
13+
- **New similar in Studio.** In the Access pillar of a read-only package, the open permission set offers "New similar". It asks for a machine name and a label (the same dialog as Setup's catalog) and posts exactly `{ name, label }` to `POST /security/permission-sets/:name/clone`. The client sends no package and no facet: the server decides where the copy lands.
14+
- **The copy opens where the answer says.** The door's answer names the package the copy was saved in. The pillar opens that package's Access pillar with the copy selected (`/studio/<package>/access?surface=permission:<copy>`), and shows the server's notice there as sent. An answer that names no package (absent or `null`) or no copy says so, and opens nothing.
15+
- **Refusals and gates.** A refusal (400, 403, 404, 409) is shown in the server's words in the dialog. Under a walled posture, or for a caller without `manage_metadata`, the action is drawn disabled with the console's own sentence for why. The base's sets stay read-only, and no override gesture is drawn on a permission set.
16+
- **One label.** The action reads "New similar" (zh: 新建类似) on Setup's catalog too, and the guidance strings that name it say the server saves the copy in the package that holds your changes. Setup's post-clone hop is unchanged.
17+
- **Access pillar keyed by package.** Like its sibling pillars, it remounts when the package changes, so opening the copy in another package's pillar no longer keeps the base's open set.
Lines changed: 22 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,22 @@
1+
---
2+
"@objectstack/service-automation": minor
3+
---
4+
5+
fix(service-automation)!: a connector stored in the environment never reads the host's files
6+
7+
A provider-bound connector saved through the metadata door (`PUT /api/v1/meta/connector/:name`, Studio), environment-wide or into a package created in the environment (`POST /api/v1/packages`, an `authored` package), had its package-relative refs resolved against the HOST's root, as if the host's own stack had declared it. A stored definition is no package's declaration and has no directory on disk, so a relative ref it names is now refused, the way an installed package's has been since #28534.
8+
9+
Clause-②: no (narrowing) — a stored connector entry (environment-wide, or in an `authored` package) is refused its package-relative file and path refs, which it was handed before against the host's root. Nothing is widened.
10+
11+
Metadata-protocol: narrowed (a stored connector's `providerConfig` may no longer name a package-relative file, and a stored stdio `mcp` transport is refused)
12+
13+
<!-- adr-0087: not-required (no-migration-prescription) No key is added, removed or renamed and the stored shape is unchanged: the same `providerConfig` is accepted at the save door, and what is refused is the host's root being read on the stored entry's behalf. No conversion can carry a file the environment never held; the refusal names the entry, its package if it has one, the ref and the remedy. -->
14+
15+
**BREAKING** behaviour narrowing for stored connectors (`@objectstack/service-automation`), shipped as `minor`.
16+
17+
- **Before.** A stored entry's package-relative refs (an `openapi` spec path, a stdio `mcp` transport's working directory) resolved against the host's root.
18+
- **Now.** Every such ref is refused for a stored entry, naming the entry, its package if it has one, and the ref: a configuration fault, so the instance is not materialized and the reconcile logs the refusal. It never fails a boot: a stored row already at rest when the platform starts is logged and skipped like one saved while it runs, and the host's own entries beside it serve. Only an entry the host's stack declares reads the host's root: one a package registration filed with the same provider binding, or one the host's own artifact carries. A stored row is never read as the host's, whatever `_packageId` its body carries.
19+
- **Unchanged.** A stored entry carrying its document inline or as an `http(s)` URL, every entry of the host's own stack, and an installed package's entries (refused as before).
20+
- **Who sees it.** A connector saved in the environment whose `openapi` spec is a package-relative path, or whose `mcp` transport is stdio.
21+
22+
**The remedy.** Carry the document inline in the definition, or name it by an `http(s)` URL; reach an MCP server from a stored entry over `http`.
Lines changed: 20 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,20 @@
1+
---
2+
"@objectstack/metadata-protocol": patch
3+
---
4+
5+
fix(metadata-protocol)!: a stored row's package in the metadata list is its column's, never its body's
6+
7+
The metadata list read (`getMetaItems`, and the execution read the automation engine binds from) served a stored row's `_packageId` from its `package_id` column only when the body carried none, so a row whose body carried another package's id was served as that package's item. The column now wins: a body stamp is replaced by the column's package, or dropped when the row is bound to no package. The save door has stripped a body's `_packageId` since #16702, so only a row written before that, or by another writer, carries one.
8+
9+
Clause-②: no (narrowing) — a stored row whose body carries a `_packageId` other than its column's is served under its column's package (or none), no longer under the body's. Nothing is widened.
10+
11+
Metadata-protocol: none
12+
13+
<!-- adr-0087: not-required (no-migration-prescription) No key is added, removed or renamed and no stored body is rewritten or refused: the row's package binding is its `package_id` column, which the read already served whenever the body carried no stamp; a body stamp that disagrees with the column was never written by the save door, so no author has anything to convert. -->
14+
15+
**BREAKING** read narrowing (`@objectstack/metadata-protocol`), shipped as `patch`.
16+
17+
- **Before.** A stored row whose body carried `_packageId` was served with the body's value, whatever its column said.
18+
- **Now.** It is served with its column's package, or with none when the row is bound to no package.
19+
- **Unchanged.** A row whose body carries no stamp, or one equal to its column, is served exactly as before.
20+
- **Who sees it.** Only a row written before the save door began stripping body stamps, or by a writer other than the save door.
Lines changed: 14 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,14 @@
1+
---
2+
'@objectstack/cloud-connection': patch
3+
---
4+
5+
fix(cloud-connection)!: install-local judges a ledger entry stored in the CLI's lifted shape through the install door's flatten, at the boot rehydrate and on the sample-data doors (#28604)
6+
7+
Clause-②: no (narrowing) — at `kernel:ready` the install-local rehydrate no longer loads, and `POST /api/v1/marketplace/install-local/:manifestId/reseed-sample-data` and `/purge-sample-data` answer 422 `OS_PROTOCOL_INCOMPATIBLE` for, a ledger entry written by an install before #28535 in the body `os package install <file>` sends (`id` and `version` lifted beside a nested `manifest`) whose nested `engines.protocol` excludes this runtime. Such an entry loaded and those doors acted on it before. A flat entry with the same range was already refused the same way. Nothing is widened.
8+
Metadata-protocol: none
9+
10+
<!-- adr-0087: not-required (no-migration-prescription) No key of any schema is added, removed, renamed or re-shaped, and no export or type changes: the rehydrate's existing ADR-0087 handshake, its ADR-0141 D5 dependency order and the sample-data doors' handshake now read a lifted ledger entry through the flatten the install door already uses, as they already read a flat entry. The refusals are the existing shared answers, so there is nothing for a migration to rewrite. -->
11+
12+
**BREAKING** boot-load and accept-set narrowing for already-ledgered install-local entries (`@objectstack/cloud-connection`): the `kernel:ready` rehydrate and the two sample-data doors, reached by a deployment where `os package install ./dist/objectstack.json` ran before #28535.
13+
14+
An install before #28535 ledgered the CLI's body unflattened, so the entry's `engines` and `dependencies` sat one level down, and the rehydrate and the sample-data doors judged `engines.protocol` at the top: an entry built for another protocol was re-registered at every boot with no error. They now read the entry through the install door's one flatten (`flattenPackageBody`), and the rehydrate orders entries by the dependencies they nest too. Such an entry gets exactly what a flat entry with the same range gets: it is not loaded, one error line names the code, the package and the remedy, the listing marks it `notLoaded`, and DELETE or a compatible reinstall still act on it. A compatible lifted entry rehydrates exactly as before, registering the body as it was ledgered.
Lines changed: 22 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,22 @@
1+
---
2+
"@objectstack/spec": minor
3+
---
4+
5+
feat(spec)!: `defineStack`, `os validate` and `os build` refuse a collection item whose body `_packageId` names a package other than the one that ships it
6+
7+
The stack schema (`ObjectStackDefinitionSchema`) now judges every item of the collections the registration door files under their package (`OWNER_STAMP_JUDGED_COLLECTIONS`: connectors, flows, views, actions, permission sets, seed `data` and the rest) and raises an issue at the item's `_packageId` when the body carries a stamp other than the package that ships it: the stack's `manifest.id` (no package without a manifest), or for a release artifact each `packages[]` body's own `id`, with a top-level item no body declares judged against `manifest.id`. `defineStack` answers `STACK_SCHEMA_INVALID` / 422; `os validate`, `os build` and the artifact door's parse report the located issue. The judge is one pure function, `judgeCollectionItemOwnerStamp` (`@objectstack/spec/kernel`), which `@objectstack/objectql`'s registration refusal calls too, so the build door and the boot cannot disagree.
8+
9+
Clause-②: no (narrowing) — the stack schema refuses a collection item whose body `_packageId` names a package other than the one that ships it, which it parsed before and the registration refused only at boot or install. Nothing is widened.
10+
11+
Metadata-protocol: narrowed (a collection item's `_packageId`, declared by `MetadataProtectionFields`, is refused at the build doors unless it is absent or names the package that ships the item)
12+
13+
<!-- adr-0087: not-required (no-migration-prescription) No key is added, removed or renamed, and no value a legitimate producer writes is refused: the registration is the only writer of `_packageId` (`os build` writes none, the package export strips every underscore key), and a stamp equal to the shipping package is still accepted. A foreign stamp exists only in a body somebody wrote by hand, which the registration door already refuses at boot or install; this moves the same refusal to where the body is written. A conversion could not tell a foreign stamp from an agreeing one without the owner that judges it, and the refusal names the item, both values and the remedy. -->
14+
15+
**BREAKING** accept-set narrowing at the build doors (`@objectstack/spec`), shipped as `minor`.
16+
17+
- **Before.** `defineStack({ manifest: { id: 'com.acme.app', … }, connectors: [{ name: 'billing', …, _packageId: 'com.acme.host' }] })` parsed, and `os validate` exited 0; `ObjectQL.registerApp` refused the stack at boot or install with `VALIDATION_ERROR` / 400 at `_packageId`.
18+
- **Now.** The parse raises an issue at `connectors.0._packageId` whose message starts "Drop `_packageId` from connector 'billing'", names the stamp and the package that ships the item, and `os validate` / `os build` exit 1. `''` is refused the same way; `null` is refused by the item schema's own type check at the same key.
19+
- **Unchanged.** An absent `_packageId` and one equal to the shipping package parse as before. Objects, object extensions, apps, picklist extensions, datasources, datasource mappings and translations are not judged at either door. A top-level copy of a `packages[]` body's item is not judged against the artifact's id. A stack's `plugins` (runtime plugin instances) are not read by the parse; the registration door judges a nested plugin's items when it registers them.
20+
- **Who sees it.** Whoever hand-writes `_packageId` on a stack item naming another package, or on any item of a stack with no `manifest`, which no build, export or install writes.
21+
22+
**The remedy.** Drop `_packageId` from the body. The package an item belongs to is the one that ships and registers it, stamped by the registration.
Lines changed: 22 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,22 @@
1+
---
2+
"@objectstack/objectql": patch
3+
---
4+
5+
refactor(objectql): the registration door's `_packageId` refusal calls the shared spec judge
6+
7+
`collectionItemOwnerStampRefusal` now asks `judgeCollectionItemOwnerStamp` (`@objectstack/spec/kernel`) for the verdict and the words, the judge the stack schema runs at `defineStack`, `os validate` and `os build`, so the build door and the boot cannot disagree. Its `VALIDATION_ERROR` / 400 envelope with `issues[]` at `_packageId`, its words, and its exemption for a stamp a registration of this process wrote onto the caller's object are unchanged.
8+
9+
Clause-②: no (narrowing) — carried by `@objectstack/spec`'s `minor` in the same change: the build doors now refuse the body `_packageId` this door already refused. This package's own accept set is unchanged. Nothing is widened.
10+
11+
Metadata-protocol: narrowed (a collection item's `_packageId` is refused at the build doors as it already was at registration)
12+
13+
<!-- adr-0087: not-required (no-migration-prescription) This package changes no key, value or verdict: the registration door refuses exactly what it refused before, in the same words. The narrowing it travels with is the build doors', declared on `@objectstack/spec`. -->
14+
15+
**BREAKING** accept-set narrowing at the build doors, declared on `@objectstack/spec`'s `minor`; this `patch` carries the shared judge's caller only.
16+
17+
- **Before.** The registration door compared a body's `_packageId` with the registering package itself.
18+
- **Now.** It calls the spec judge, which also runs at the build doors.
19+
- **Unchanged.** Every verdict and every word at `registerApp`, the seed datasets, the `viewItems:` channel and the `manifest` service's residual registration.
20+
- **Who sees it.** Nobody at this door; an author now hears the same refusal at `os validate`.
21+
22+
**The remedy.** Drop `_packageId` from the body.
Lines changed: 16 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,16 @@
1+
---
2+
'@objectstack/spec': patch
3+
---
4+
5+
fix(spec): the datasource read redactor covers the declared `ssl` secret members
6+
7+
Clause-②: no (narrowing)
8+
Metadata-protocol: none
9+
10+
<!-- adr-0087: not-required (no-migration-prescription) No metadata moves: no spec key, authorable spelling, export or stored shape is removed, renamed or re-shaped, and no authored value is refused, so there is nothing for `objectstack migrate meta` to rewrite. What narrows is a read answer: the metadata read exits serve a datasource without `ssl.key` and name it in `redactedKeys`, the posture they already apply to `config` credentials. The stored row keeps the key for the connect path, and the generic write-path carry-forward keeps it on a save that omits it. No exported declaration is removed or narrowed (not runtime-interface-only or type-surface-only); the package publishes (not unpublished); no ADR-0087 id is named or touched (not registered or already-registered). -->
11+
12+
**BREAKING** (a read-answer narrowing), shipped as `patch`: a non-protocol narrowing takes `patch` and keeps this banner and its ADR-0087 disposition.
13+
14+
`DatasourceSchema.ssl.key` (the client private key) is now marked a credential value (`.meta({ credential: 'value' })`), and the built-in `datasource` read redactor applies the schema's credential markers beside its `config` rule. Every metadata read exit that resolves the `datasource` redactor (`/meta` item, list and layered reads, and the stored-row seam behind the `sys_metadata` data door) drops `ssl.key` from the served item and lists `ssl.key` in `redactedKeys`.
15+
16+
**Unchanged.** `ssl.enabled`, `ssl.rejectUnauthorized`, `ssl.ca` and `ssl.cert` are public material and stay served. A datasource with no `ssl` block, or with an empty `key`, is served exactly as stored. The stored row is never rewritten, and a save that sends back the served body keeps the stored key.
Lines changed: 12 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,12 @@
1+
---
2+
'@objectstack/plugin-approvals': patch
3+
---
4+
5+
fix(plugin-approvals): an approver that resolves the same person twice opens one slot for them, so one signature counts once (#28684)
6+
7+
Clause-②: no
8+
Metadata-protocol: none
9+
10+
An approval step builds its slate one approver entry at a time, and an entry could name the same person more than once. An `expression` approver yielding `['u2', 'u2']` (for example a lead picking the same reviewer twice for the next step), a `field` approver holding `'u2,u2'`, two ids whose out-of-office delegation routes to the same delegate, or a `resolveAs` expression returning two departments that share a member each opened a slot per occurrence. The request then listed that person twice as pending, and under `behavior: 'per_group'` their one approval counted twice toward `minApprovals`, so a group could finalize on a single signature.
11+
12+
Now each approver entry names each resolved person once, after out-of-office delegation, in first-appearance order. Every group the person resolved into is still recorded, and a person tagged into the same group twice (also when two approver entries name them under one group) counts once in that group's tally. Two approver entries that both name one person still give one slot per entry, as before.

0 commit comments

Comments
 (0)