Concurrent publishes for the same (namespace_id, slug, owner_id) or
(skill_id, version) coordinate both pass the check-then-create reads and
race on the database unique constraints. The losing request surfaced an
unhandled DataIntegrityViolationException as HTTP 500.
Translate the constraint violation at both insert points into a
deterministic DomainBadRequestException (error.skill.publish.concurrentConflict),
matching the existing idiom in LabelDefinitionService/ReviewService/
PromotionService. No same-transaction re-read is attempted, so the losing
publish rolls back cleanly and returns a retryable conflict instead of a 500.
Add the i18n key (en/zh) and two unit tests covering the skill-insert and
version-insert races.
Closes#617
Signed-off-by: FenjuFu <92919259+FenjuFu@users.noreply.github.com>
audit_log.detail_json is a JSONB column, so the value has to be valid
JSON or the insert fails. It was built by string concatenation at every
call site, with three inconsistent levels of escaping: none at all in
ClawHubCompatAppService, DeviceAuthWebController, LabelAdminAppService
and SkillLabelAppService; quotes only in ReviewPortalAppService,
SkillLifecycleAppService, NamespaceGovernanceService and
SkillGovernanceService; quotes and backslashes in
PromotionPortalAppService.escapeJson.
None of the three escapes control characters, which JSON forbids raw
inside a string. A reviewer pressing Enter in a review comment therefore
produced a payload PostgreSQL rejects, and because the audit write
happens after the domain mutation, the review was already approved when
the request returned 500.
Add AuditDetail, which renders the payload through Jackson, and route
all 29 construction sites through it. 17 of those interpolate a string
value and are the actual defect surface; the numeric and constant ones
are converted too so there is one way to build audit detail and no
hand-rolled example left to copy.
SkillHardDeleteService.toAuditPayload already did this correctly with a
LinkedHashMap and an ObjectMapper; AuditDetail is that shape extracted.
The service itself is left alone rather than changing its constructor
signature for no behavior gain.
Output is byte-identical for values that were already escaped correctly,
so the existing exact-string assertions in AdminSearchControllerTest and
PromotionPortalAppServiceTest are unchanged. null still means "no
detail": the builder returns null rather than {} when no field is set.
Addresses the JSON half of #615. The transaction half -- the domain
mutation and the audit write not sharing one transaction -- is a
separate design decision about whether an audit failure should roll back
a review, and is not bundled here.
Signed-off-by: FenjuFu <fufenjupku@gmail.com>
Use a bounded safe YAML constructor, remove the polynomial placeholder regex, keep public label access GET-only with CSRF protection, validate CLI callbacks as loopback URL objects, and use cryptographic UUIDs for E2E identities.
Signed-off-by: FenjuFu <92919259+FenjuFu@users.noreply.github.com>
* feat(api): return skill labels from the skill listing endpoints on request
Skill labels were reachable only one skill at a time, through
/api/{v1,web}/skills/{namespace}/{slug}/labels, so a client rendering a list had
to issue a follow-up request per row.
Add includeLabels=true to GET /api/v1/skills and GET /api/web/skills. The labels
array is populated only when the parameter is set and left out of the payload
otherwise, so existing responses are byte-identical.
Labels for the whole page are resolved by SkillLabelProjectionService in three
queries — assignments, definitions, translations — rather than three per skill.
Closes#710
Signed-off-by: FenjuFu <fufenjupku@gmail.com>
* fix(api): use include parameter for skill labels
Signed-off-by: XiaoSeS <87064762+XiaoSeS@users.noreply.github.com>
* test(api): reject unsupported include before search
Signed-off-by: XiaoSeS <87064762+XiaoSeS@users.noreply.github.com>
---------
Signed-off-by: FenjuFu <fufenjupku@gmail.com>
Signed-off-by: XiaoSeS <87064762+XiaoSeS@users.noreply.github.com>
Co-authored-by: XiaoSeS <87064762+XiaoSeS@users.noreply.github.com>
* fix(security): publish scan task after transaction commit
SecurityScanService.triggerScan is @Transactional but published the Redis
Stream scan task inline, before the transaction committed. The stream
consumer could receive the task before the skill_version / security_audit
rows were visible, fail with "SkillVersion not found" / "SecurityAudit not
found", exhaust its immediate retries while the publishing transaction was
still open, and leave the committed version stuck in SCANNING.
Defer the publish to an afterCommit transaction synchronization so the
consumer only ever sees the task once the rows are committed and visible; on
rollback the task is never published. Falls back to an inline publish when
called outside a transaction.
Closes#612
Signed-off-by: FenjuFu <fufenjupku@gmail.com>
* test(security): cover scan task after-commit publishing
Signed-off-by: XiaoSeS <87064762+XiaoSeS@users.noreply.github.com>
* refactor(security): hide scan publish transaction callback
Signed-off-by: XiaoSeS <87064762+XiaoSeS@users.noreply.github.com>
---------
Signed-off-by: FenjuFu <fufenjupku@gmail.com>
Signed-off-by: XiaoSeS <87064762+XiaoSeS@users.noreply.github.com>
Co-authored-by: XiaoSeS <87064762+XiaoSeS@users.noreply.github.com>
A deployment that stands up its own organisation-wide namespace — to use
instead of the built-in global one — finds it invisible to everybody. The
namespace listing only returns namespaces the caller belongs to, and the
only thing that ever added members automatically was hard-wired to the slug
"global".
Make that list a setting. namespace.default-membership holds the slugs every
newly activated account is enrolled in, defaulting to ["global"], which is
what every deployment did before. Publishing only requires membership of any
role, so being enrolled is enough to publish there; no extra grant needed.
GlobalNamespaceMembershipService becomes DefaultNamespaceMembershipService,
since it no longer means one specific namespace.
Where the strictness sits:
- Saving validates every slug resolves to an ACTIVE namespace, so a typo
fails at the moment an administrator makes it.
- Enrolling tolerates a slug that no longer resolves: it logs and skips.
A namespace that was deleted or renamed must not cost somebody their
login.
Adding a namespace to the list after people have signed up leaves them out,
the same trap the personal-namespace work hit, so this ships with the same
preview-then-apply backfill.
Verified against a real PostgreSQL, end to end: create a namespace, reject an
unknown slug, save with whitespace and duplicates and see them normalised,
preview, apply, re-preview showing nothing left, and a fresh registration
landing in the global namespace, the new shared one, and its own personal one
at once.
