Add the Greptile and ReviewBench benchmark data

greptile/README.md

@@ -0,0 +1,23 @@
+# Greptile benchmark
+
+Perch's runs on [Greptile's benchmark](https://www.greptile.com/benchmarks): 50 pull requests, each carrying one known bug, with the golden comments from [Martian's code review benchmark](https://github.com/withmartian/code-review-benchmark). The scripts that score these files are in [`.benchmarks/greptile`](https://github.com/lakeday-org/perch/tree/main/.benchmarks/greptile) in lakeday-org/perch. Its README describes the runs and how they are scored.
+
+These files are scored by lakeday-org/perch at commit `6b33c162a778f56e27ae043f57f4d52cf30b9456`, from [lakeday-org/perch#421](https://github.com/lakeday-org/perch/pull/421). Download them into a perch checkout, where the scripts read them:
+
+```sh
+hf download perchscan/benchmark-results --repo-type dataset --include 'greptile/*' --local-dir .benchmarks/hf
+```
+
+| File | Contents |
+|---|---|
+| `manifest.json` | Each pull request in its [ai-code-review-evaluation](https://github.com/ai-code-review-evaluation) fork, with its base and head commits and Martian's golden comments |
+| `known.json` | Each pull request's known bug, in Greptile's words, and the functions that hold it |
+| `truth/<repo>.json` | The functions each golden comment is about, for Cal.com, Discourse, Grafana, Keycloak and Sentry |
+| `results/oai2q.jsonl.gz` | perch#418 on OpenAI Decisions, one line a pull request |
+| `results/oai407.jsonl.gz` | perch#407 on OpenAI Decisions |
+| `results/jev048.jsonl.gz` | The released perch 0.4.8 on Jev (`jev-latest`) |
+| `results/jev.jsonl.gz` | perch 0.4.7 on Jev |
+| `results/oai.jsonl.gz` | perch 0.4.7 on OpenAI Decisions |
+| `summary.json` | Each run's scores at the 60%, 70% and 80% cutoffs, from `score.mjs` |
+
+Martian's golden comments are [MIT-licensed](https://github.com/withmartian/code-review-benchmark). Source code is licensed according to its repository.

greptile/known.json

@@ -0,0 +1,565 @@
+[
+ {
+  "repo": "sentry",
+  "pr": 1,
+  "bug": "Pagination for audit logs imports a nonexistent OptimizedCursorPaginator / negative slice offsets",
+  "methods": [
+   "src/sentry/api/paginator.py::OptimizedCursorPaginator.get_result"
+  ],
+  "file": "src/sentry/api/paginator.py",
+  "line": 880,
+  "why": "Matches golden 0/3: the negative-offset branch assigns start_offset = cursor.offset and slices the queryset with it (line 882); the import named in the description is real at head."
+ },
+ {
+  "repo": "sentry",
+  "pr": 2,
+  "bug": "Negative offset cursor bypasses pagination bounds during span buffer insert",
+  "methods": [
+   "src/sentry/api/paginator.py::OptimizedCursorPaginator.get_result"
+  ],
+  "file": "src/sentry/api/paginator.py",
+  "line": 880,
+  "why": "The described negative-offset bypass is the paginator change, not the span buffer insert: the span-buffer files (add-buffer.lua, buffer.py, factory.py) have no cursor or offset logic; the defect is start_offset = cursor.offset fed to queryset slicing."
+ },
+ {
+  "repo": "sentry",
+  "pr": 3,
+  "bug": "Upsampled error count skips sample_rate of 0.0 because it is treated as falsy",
+  "methods": [
+   "src/sentry/testutils/factories.py::_set_sample_rate_from_error_sampling"
+  ],
+  "file": "src/sentry/testutils/factories.py",
+  "line": 353,
+  "why": "Same as golden 0: the truthiness check `if client_sample_rate:` drops a 0.0 sample rate."
+ },
+ {
+  "repo": "sentry",
+  "pr": 4,
+  "bug": "Null reference when github_authenticated_user state is missing (OAuth change)",
+  "methods": [
+   "src/sentry/integrations/github/integration.py::GitHubInstallation.dispatch"
+  ],
+  "file": "src/sentry/integrations/github/integration.py",
+  "line": 502,
+  "why": "Same as golden 0: the github_authenticated_user state is fetched and compared with no presence check."
+ },
+ {
+  "repo": "sentry",
+  "pr": 5,
+  "bug": "Replays bulk delete changes the error response format, breaking clients",
+  "methods": [
+   "src/sentry/issues/endpoints/browser_reporting_collector.py::BrowserReportingCollectorEndpoint.post"
+  ],
+  "file": "src/sentry/issues/endpoints/browser_reporting_collector.py",
+  "line": 115,
+  "why": "The description does not match the replays bulk-delete code (replays/usecases/delete.py only adds a None max_segment_id guard); the error response change in this PR is the browser reporting collector, which now answers an invalid report with a 422 JSON body {'error', 'details'} where the base code raised a TypeError (500) from BrowserReport(**report)."
+ },
+ {
+  "repo": "sentry",
+  "pr": 6,
+  "bug": "Inconsistent metric tag names ('shard' vs 'shards') in span buffer monitoring",
+  "methods": [
+   "src/sentry/spans/consumers/process/flusher.py::SpanFlusher.main"
+  ],
+  "file": "src/sentry/spans/consumers/process/flusher.py",
+  "line": 199,
+  "why": "Same as golden 0: 'shards' tag on wait_produce versus 'shard' on the other flusher metrics."
+ },
+ {
+  "repo": "sentry",
+  "pr": 7,
+  "bug": "Dataclass uses a shared mutable default for a timestamp",
+  "methods": [
+   "src/sentry/integrations/services/assignment_source.py::<top-level>"
+  ],
+  "file": "src/sentry/integrations/services/assignment_source.py",
+  "line": 18,
+  "why": "Same as golden 0: the dataclass default timezone.now() is evaluated once at class definition and shared by all instances."
+ },
+ {
+  "repo": "sentry",
+  "pr": 8,
+  "bug": "Incident creation reads a stale config variable instead of the updated one",
+  "methods": [
+   "src/sentry/monitors/logic/incident_occurrence.py::get_monitor_environment_context"
+  ],
+  "file": "src/sentry/monitors/logic/incident_occurrence.py",
+  "line": 168,
+  "why": "Same as golden 0: the context returns the original monitor.config instead of the updated config copy."
+ },
+ {
+  "repo": "sentry",
+  "pr": 9,
+  "bug": "Queue-based parallelism mishandles the queue.ShutDown exception",
+  "methods": [
+   "src/sentry/remote_subscriptions/consumers/queue_consumer.py::OrderedQueueWorker.run"
+  ],
+  "file": "src/sentry/remote_subscriptions/consumers/queue_consumer.py",
+  "line": 142,
+  "why": "The second except queue.ShutDown wraps result_processor, which never raises it, so it is dead code; ShutDown only comes from work_queue.get() (line 132), and FixedQueuePool.shutdown sets worker.shutdown = True before q.shutdown(immediate=False), so workers leave the loop without draining queued items. It is unclear which part the description means; line 142 is the misplaced handler."
+ },
+ {
+  "repo": "sentry",
+  "pr": 10,
+  "bug": "Occurrence hook in stateful detector is an incomplete stub (only pass)",
+  "methods": [
+   "src/sentry/incidents/grouptype.py::<top-level>"
+  ],
+  "file": "src/sentry/incidents/grouptype.py",
+  "line": 11,
+  "why": "Same as golden 0: MetricAlertDetectorHandler is a stub containing only `pass`."
+ },
+ {
+  "repo": "grafana",
+  "pr": 1,
+  "bug": "Race condition in device creation/update (anonymous device limit), CreateOrUpdateDevice",
+  "methods": [
+   "pkg/services/anonymous/anonimpl/anonstore/database.go::AnonDBStore.CreateOrUpdateDevice"
+  ],
+  "file": "pkg/services/anonymous/anonimpl/anonstore/database.go",
+  "line": 110,
+  "why": "Check-then-insert race: the device count and the upsert are separate, unsynchronized DB operations."
+ },
+ {
+  "repo": "grafana",
+  "pr": 2,
+  "bug": "Authz cache entries lack expiration, causing permanent permission denials",
+  "methods": [
+   "pkg/services/authz/rbac/service.go::Service.Check",
+   "pkg/services/authz/rbac/service.go::NewService"
+  ],
+  "file": "pkg/services/authz/rbac/service.go",
+  "line": 117,
+  "why": "Denials written to permDenialCache (line 154) short-circuit Check at line 117 and are never invalidated when permissions change; note the cache does get shortCacheTTL (30s) in NewService line 85, so 'no expiration' is overstated and the defensible defect is the stale-denial short-circuit in Check."
+ },
+ {
+  "repo": "grafana",
+  "pr": 3,
+  "bug": "Undefined endpoint constants after middleware rename cause compile errors",
+  "methods": [
+   "pkg/services/pluginsintegration/clientmiddleware/contextual_logger_middleware.go::ContextualLoggerMiddleware.QueryData",
+   "pkg/services/pluginsintegration/clientmiddleware/contextual_logger_middleware.go::ContextualLoggerMiddleware.CallResource",
+   "pkg/services/pluginsintegration/clientmiddleware/contextual_logger_middleware.go::ContextualLoggerMiddleware.CheckHealth",
+   "pkg/services/pluginsintegration/clientmiddleware/contextual_logger_middleware.go::ContextualLoggerMiddleware.CollectMetrics"
+  ],
+  "file": "pkg/services/pluginsintegration/clientmiddleware/contextual_logger_middleware.go",
+  "line": 40,
+  "why": "Unsure: the endpoint constants used by the new middleware (lines 40, 45, 50, 55) are in fact defined in clientmiddleware/utils.go at head, so no compile error is visible; these uses are the closest match to the described defect."
+ },
+ {
+  "repo": "grafana",
+  "pr": 4,
+  "bug": "Double interpolation risk in query processing",
+  "methods": [
+   "public/app/plugins/datasource/loki/querySplitting.ts::runSplitQuery",
+   "public/app/plugins/datasource/loki/shardQuerySplitting.ts::runShardSplitQuery"
+  ],
+  "file": "public/app/plugins/datasource/loki/querySplitting.ts",
+  "line": 299,
+  "why": "Queries are interpolated with applyTemplateVariables before splitting, then interpolated again when datasource.runQuery calls super.query (DataSourceWithBackend applies template variables); same pattern in shardQuerySplitting.ts line 52."
+ },
+ {
+  "repo": "grafana",
+  "pr": 5,
+  "bug": "Missing key prop causes React rendering issues in notification rules",
+  "methods": [
+   "public/app/features/alerting/unified/rule-list/FilterView.tsx::FilterViewResults"
+  ],
+  "file": "public/app/features/alerting/unified/rule-list/FilterView.tsx",
+  "line": 157,
+  "why": "GrafanaRuleListItem in the mapped list is missing its key prop."
+ },
+ {
+  "repo": "grafana",
+  "pr": 6,
+  "bug": "Wrong metrics recording methods produce misleading performance data",
+  "methods": [
+   "pkg/apiserver/rest/dualwriter_mode3.go::DualWriterMode3.Create",
+   "pkg/apiserver/rest/dualwriter_mode3.go::DualWriterMode3.Update",
+   "pkg/apiserver/rest/dualwriter_mode3.go::DualWriterMode3.DeleteCollection"
+  ],
+  "file": "pkg/apiserver/rest/dualwriter_mode3.go",
+  "line": 45,
+  "why": "Wrong recorder used: storage failures recorded via recordLegacyDuration in Create (45) and Update (129), and legacy DeleteCollection recorded via recordStorageDuration (166)."
+ },
+ {
+  "repo": "grafana",
+  "pr": 7,
+  "bug": "Incorrect error log level in database optimizations",
+  "methods": [
+   "pkg/services/annotations/annotationsimpl/xorm_store.go::xormRepositoryImpl.CleanAnnotations",
+   "pkg/services/annotations/annotationsimpl/xorm_store.go::xormRepositoryImpl.CleanOrphanedAnnotationTags"
+  ],
+  "file": "pkg/services/annotations/annotationsimpl/xorm_store.go",
+  "line": 534,
+  "why": "Error log level used for routine debug information in the annotation cleanup batches."
+ },
+ {
+  "repo": "grafana",
+  "pr": 8,
+  "bug": "Possible deadlock during concurrent annotation deletion",
+  "methods": [],
+  "file": null,
+  "line": null,
+  "why": "Unsure: this PR only changes GetWebAssets locking (no annotation deletion code, and the RLock/Lock sequence cannot deadlock); the description matches PR 7's annotation cleanup, so there is no location in this PR."
+ },
+ {
+  "repo": "grafana",
+  "pr": 9,
+  "bug": "enableSqlExpressions always returns false, disabling SQL functionality",
+  "methods": [
+   "pkg/expr/reader.go::enableSqlExpressions"
+  ],
+  "file": "pkg/expr/reader.go",
+  "line": 195,
+  "why": "enableSqlExpressions returns false whether or not the feature flag is enabled."
+ },
+ {
+  "repo": "grafana",
+  "pr": 10,
+  "bug": "Race condition in cache locking (unified storage)",
+  "methods": [
+   "pkg/storage/unified/search/bleve.go::bleveBackend.BuildIndex"
+  ],
+  "file": "pkg/storage/unified/search/bleve.go",
+  "line": 137,
+  "why": "Narrowing the cacheMu lock in BuildIndex to the cache write leaves index construction unsynchronized."
+ },
+ {
+  "repo": "cal.com",
+  "pr": 2,
+  "bug": "Async callbacks in forEach cause unhandled promise rejections (appStore import)",
+  "methods": [
+   "packages/app-store/vital/lib/reschedule.ts::Reschedule",
+   "packages/app-store/wipemycalother/lib/reschedule.ts::Reschedule",
+   "packages/trpc/server/routers/viewer/bookings.tsx::requestReschedule",
+   "packages/features/bookings/lib/handleCancelBooking.ts::handler"
+  ],
+  "file": "packages/app-store/vital/lib/reschedule.ts",
+  "line": 125,
+  "why": "forEach(async (bookingRef) => { await getCalendar(...) }) leaves the promises unawaited, causing unhandled rejections (same as golden 1; also wipemycalother reschedule.ts:125, bookings.tsx:553, handleCancelBooking.ts:460)."
+ },
+ {
+  "repo": "cal.com",
+  "pr": 3,
+  "bug": "2FA backup codes are not invalidated after use",
+  "methods": [
+   "packages/features/auth/lib/next-auth-options.ts::authorize"
+  ],
+  "file": "packages/features/auth/lib/next-auth-options.ts",
+  "line": 148,
+  "why": "A used backup code is invalidated only by a non-atomic read-modify-write of the encrypted list, so it is not reliably invalidated after use (concurrent reuse, as in golden 3); the disable endpoint clears all codes, so login is the only place this applies."
+ },
+ {
+  "repo": "cal.com",
+  "pr": 4,
+  "bug": "Null reference when destinationCalendar host array is empty",
+  "methods": [
+   "packages/core/EventManager.ts::EventManager.create",
+   "packages/app-store/googlecalendar/lib/CalendarService.ts::GoogleCalendarService.createEvent",
+   "packages/app-store/googlecalendar/lib/CalendarService.ts::GoogleCalendarService.updateEvent"
+  ],
+  "file": "packages/core/EventManager.ts",
+  "line": 119,
+  "why": "const [mainHostDestinationCalendar] = evt.destinationCalendar ?? [] is undefined for an empty array, and mainHostDestinationCalendar.integration is then read without a guard; the Google createEvent/updateEvent destructuring named in golden 0 is listed too."
+ },
+ {
+  "repo": "cal.com",
+  "pr": 5,
+  "bug": "SQL injection risk in raw Prisma.sql query construction (InsightsBookingService)",
+  "methods": [
+   "packages/lib/server/service/insightsBooking.ts::InsightsBookingService.buildFilterConditions",
+   "packages/lib/server/service/insightsBooking.ts::InsightsBookingService.buildAuthorizationConditions",
+   "packages/lib/server/service/insightsBooking.ts::InsightsBookingService.buildOrgAuthorizationCondition",
+   "packages/lib/server/service/insightsBooking.ts::InsightsBookingService.buildTeamAuthorizationCondition"
+  ],
+  "file": "packages/lib/server/service/insightsBooking.ts",
+  "line": 106,
+  "why": "WHERE fragments are now built as raw Prisma.sql strings from filter and option values (first at line 106); Prisma.sql binds the interpolated values as parameters, so the injection risk is nominal, and the location is ambiguous across all four build* methods."
+ },
+ {
+  "repo": "cal.com",
+  "pr": 6,
+  "bug": "Workflow reminders not cleaned up from DB when immediateDelete is true",
+  "methods": [
+   "packages/features/ee/workflows/lib/reminders/emailReminderManager.ts::deleteScheduledEmailReminder"
+  ],
+  "file": "packages/features/ee/workflows/lib/reminders/emailReminderManager.ts",
+  "line": 213,
+  "why": "With immediateDelete true, the function returns after the SendGrid cancel and never removes the WorkflowReminder from the DB."
+ },
+ {
+  "repo": "cal.com",
+  "pr": 7,
+  "bug": "End time computed from slotStartTime instead of slotEndTime",
+  "methods": [
+   "packages/trpc/server/routers/viewer/slots.ts::checkIfIsAvailable"
+  ],
+  "file": "packages/trpc/server/routers/viewer/slots.ts",
+  "line": 142,
+  "why": "const end = slotStartTime.hour() * 60 + slotStartTime.minute() uses the slot start time where the slot end time is needed."
+ },
+ {
+  "repo": "cal.com",
+  "pr": 8,
+  "bug": "Timing attack from direct string comparison in OAuth credential sync",
+  "methods": [
+   "apps/web/pages/api/webhook/app-credential.ts::handler"
+  ],
+  "file": "apps/web/pages/api/webhook/app-credential.ts",
+  "line": 25,
+  "why": "The webhook secret header is compared to CALCOM_WEBHOOK_SECRET with !==, a non-constant-time comparison open to timing attacks."
+ },
+ {
+  "repo": "cal.com",
+  "pr": 9,
+  "bug": "OR condition in SMS reminder logic deletes all workflow reminders",
+  "methods": [
+   "packages/features/ee/workflows/api/scheduleSMSReminders.ts::handler"
+  ],
+  "file": "packages/features/ee/workflows/api/scheduleSMSReminders.ts",
+  "line": 38,
+  "why": "The retryCount OR condition in deleteMany is not limited to SMS reminders, so it deletes workflow reminders of every method."
+ },
+ {
+  "repo": "cal.com",
+  "pr": 10,
+  "bug": "Email blacklist is case-sensitive, allowing bypass (guest management)",
+  "methods": [
+   "packages/trpc/server/routers/viewer/bookings/addGuests.handler.ts::addGuestsHandler"
+  ],
+  "file": "packages/trpc/server/routers/viewer/bookings/addGuests.handler.ts",
+  "line": 77,
+  "why": "The guest email is compared case-sensitively against the lowercased blacklist, allowing a bypass."
+ },
+ {
+  "repo": "cal.com",
+  "pr": 11,
+  "bug": "Unreliable updatedAt field produces inaccurate calendar cache status",
+  "methods": [
+   "packages/app-store/googlecalendar/lib/CalendarService.ts::GoogleCalendarService.fetchAvailabilityAndSetCache",
+   "packages/lib/server/repository/selectedCalendar.ts::SelectedCalendarRepository.updateManyByCredentialId"
+  ],
+  "file": "packages/app-store/googlecalendar/lib/CalendarService.ts",
+  "line": 1024,
+  "why": "The updatedAt refresh relies on an updateMany with empty data, so the timestamp shown as cache status is unreliable (same as golden 0)."
+ },
+ {
+  "repo": "keycloak",
+  "pr": 1,
+  "bug": "Passkey re-auth calls ConditionalPasskeysEnabled() without UserModel",
+  "methods": [
+   "services/src/main/java/org/keycloak/authentication/authenticators/browser/UsernameForm.java::UsernameForm.authenticate"
+  ],
+  "file": "services/src/main/java/org/keycloak/authentication/authenticators/browser/UsernameForm.java",
+  "line": 47,
+  "why": "Re-authentication check calls isConditionalPasskeysEnabled() without the UserModel parameter it requires (should pass context.getUser())."
+ },
+ {
+  "repo": "keycloak",
+  "pr": 2,
+  "bug": "Recursive caching call uses session instead of delegate (getForLogin)",
+  "methods": [
+   "model/infinispan/src/main/java/org/keycloak/models/cache/infinispan/idp/InfinispanIdentityProviderStorageProvider.java::InfinispanIdentityProviderStorageProvider.getForLogin"
+  ],
+  "file": "model/infinispan/src/main/java/org/keycloak/models/cache/infinispan/idp/InfinispanIdentityProviderStorageProvider.java",
+  "line": 246,
+  "why": "getForLogin loads each cached IdP via session.identityProviders().getById (the caching layer itself) rather than idpDelegate.getById."
+ },
+ {
+  "repo": "keycloak",
+  "pr": 3,
+  "bug": "Returns default keystore provider instead of BouncyCastle (AuthzClientCryptoProvider)",
+  "methods": [
+   "authz/client/src/main/java/org/keycloak/authorization/client/util/crypto/AuthzClientCryptoProvider.java::AuthzClientCryptoProvider.getBouncyCastleProvider"
+  ],
+  "file": "authz/client/src/main/java/org/keycloak/authorization/client/util/crypto/AuthzClientCryptoProvider.java",
+  "line": 63,
+  "why": "Returns KeyStore.getInstance(KeyStore.getDefaultType()).getProvider() (default keystore provider) instead of a BouncyCastle provider."
+ },
+ {
+  "repo": "keycloak",
+  "pr": 4,
+  "bug": "Incorrect method call for exit codes in rolling-updates framework",
+  "methods": [
+   "quarkus/runtime/src/main/java/org/keycloak/quarkus/runtime/cli/command/UpdateCompatibilityCheck.java::UpdateCompatibilityCheck.run",
+   "quarkus/runtime/src/main/java/org/keycloak/quarkus/runtime/cli/command/UpdateCompatibilityMetadata.java::UpdateCompatibilityMetadata.run"
+  ],
+  "file": "quarkus/runtime/src/main/java/org/keycloak/quarkus/runtime/cli/command/UpdateCompatibilityCheck.java",
+  "line": 48,
+  "why": "Exit code for the disabled rolling-updates feature is returned with picocli.exit(...), which hard-exits via System.exit; same call duplicated in UpdateCompatibilityMetadata.run."
+ },
+ {
+  "repo": "keycloak",
+  "pr": 5,
+  "bug": "Inconsistent feature flag causes orphaned permissions (Client scopes)",
+  "methods": [
+   "services/src/main/java/org/keycloak/services/resources/admin/permissions/AdminPermissions.java::AdminPermissions.registerListener.onEvent",
+   "services/src/main/java/org/keycloak/services/resources/admin/permissions/AdminPermissions.java::AdminPermissions.registerListener"
+  ],
+  "file": "services/src/main/java/org/keycloak/services/resources/admin/permissions/AdminPermissions.java",
+  "line": 77,
+  "why": "Permission cleanup listener checks the V1 ADMIN_FINE_GRAINED_AUTHZ flag while the rest of the code uses V2, leaving orphaned permissions when only V2 is on."
+ },
+ {
+  "repo": "keycloak",
+  "pr": 6,
+  "bug": "Incorrect permission check in canManage() (Groups scopes)",
+  "methods": [
+   "services/src/main/java/org/keycloak/services/resources/admin/permissions/GroupPermissionsV2.java::GroupPermissionsV2.canManage"
+  ],
+  "file": "services/src/main/java/org/keycloak/services/resources/admin/permissions/GroupPermissionsV2.java",
+  "line": 70,
+  "why": "canManage() accepts the VIEW scope as well as MANAGE when falling back to the all-groups permission."
+ },
+ {
+  "repo": "keycloak",
+  "pr": 7,
+  "bug": "Lithuanian translation files contain Italian text",
+  "methods": [],
+  "file": "themes/src/main/resources-community/theme/base/account/messages/messages_lt.properties",
+  "line": 101,
+  "why": "Translation file (no method): Lithuanian totpStep1 set to Italian text; also login/messages/messages_lt.properties line 71."
+ },
+ {
+  "repo": "keycloak",
+  "pr": 8,
+  "bug": "Wrong parameter in null check (grantType vs. rawTokenId)",
+  "methods": [
+   "services/src/main/java/org/keycloak/protocol/oidc/encode/AccessTokenContext.java::AccessTokenContext.AccessTokenContext"
+  ],
+  "file": "services/src/main/java/org/keycloak/protocol/oidc/encode/AccessTokenContext.java",
+  "line": 73,
+  "why": "Null check passes grantType where rawTokenId was intended, so a null rawTokenId is accepted."
+ },
+ {
+  "repo": "keycloak",
+  "pr": 9,
+  "bug": "Unsafe raw List deserialization in recovery key support",
+  "methods": [
+   "testsuite/integration-arquillian/servers/auth-server/services/testsuite-providers/src/main/java/org/keycloak/testsuite/federation/BackwardsCompatibilityUserStorage.java::BackwardsCompatibilityUserStorage.getCredentials",
+   "testsuite/integration-arquillian/servers/auth-server/services/testsuite-providers/src/main/java/org/keycloak/testsuite/federation/BackwardsCompatibilityUserStorage.java::BackwardsCompatibilityUserStorage.isValid"
+  ],
+  "file": "testsuite/integration-arquillian/servers/auth-server/services/testsuite-providers/src/main/java/org/keycloak/testsuite/federation/BackwardsCompatibilityUserStorage.java",
+  "line": 238,
+  "why": "JsonSerialization.readValue(..., List.class) yields a raw List passed unchecked to createFromValues(List<String>...); isValid repeats it with a raw `List generatedKeys` at line 332-334."
+ },
+ {
+  "repo": "keycloak",
+  "pr": 10,
+  "bug": "Missing null check causes NullPointerException in concurrent group access",
+  "methods": [
+   "model/infinispan/src/main/java/org/keycloak/models/cache/infinispan/GroupAdapter.java::GroupAdapter.getSubGroupsCount"
+  ],
+  "file": "model/infinispan/src/main/java/org/keycloak/models/cache/infinispan/GroupAdapter.java",
+  "line": 275,
+  "why": "Concurrent-deletion path in getSubGroupsCount; this PR's head equals upstream commit 332c9b6e so nothing was injected, and the closest match is this null return (sibling getSubGroupsStream overloads at 256/262/268 still call modelSupplier.get() unchecked)."
+ },
+ {
+  "repo": "discourse",
+  "pr": 1,
+  "bug": "Method overwriting causes parameter mismatch (image downsizing)",
+  "methods": [
+   "app/models/optimized_image.rb::OptimizedImage.downsize#2",
+   "app/models/optimized_image.rb::OptimizedImage.downsize"
+  ],
+  "file": "app/models/optimized_image.rb",
+  "line": 149,
+  "why": "Same defect as golden 0: the second downsize definition overwrites the max_width/max_height one, so the parameter lists no longer match the callers."
+ },
+ {
+  "repo": "discourse",
+  "pr": 2,
+  "bug": "Nil reference to a nonexistent TopicUser (per-topic unsubscribe)",
+  "methods": [
+   "app/controllers/topics_controller.rb::TopicsController.unsubscribe"
+  ],
+  "file": "app/controllers/topics_controller.rb",
+  "line": 107,
+  "why": "Same as golden 0: unsubscribe dereferences tu without checking that the TopicUser record exists."
+ },
+ {
+  "repo": "discourse",
+  "pr": 3,
+  "bug": "BlockedEmail.should_block? writes to the DB during a read",
+  "methods": [
+   "app/models/blocked_email.rb::BlockedEmail.should_block?"
+  ],
+  "file": "app/models/blocked_email.rb",
+  "line": 14,
+  "why": "Same as golden 0: should_block? saves match statistics during what callers use as a read-only check."
+ },
+ {
+  "repo": "discourse",
+  "pr": 4,
+  "bug": "SSRF via open(url) without validation (embed URLs)",
+  "methods": [
+   "app/jobs/scheduled/poll_feed.rb::Jobs.PollFeed.poll_feed"
+  ],
+  "file": "app/jobs/scheduled/poll_feed.rb",
+  "line": 29,
+  "why": "Matches golden 0: open(SiteSetting.feed_polling_url) with no validation. TopicEmbed.import_remote also calls open(url) on the embed URL (topic_embed.rb line 48), but TopicRetriever.invalid_host? checks that host first."
+ },
+ {
+  "repo": "discourse",
+  "pr": 5,
+  "bug": "float: left mixed with flexbox causes header layout issues",
+  "methods": [],
+  "file": "app/assets/stylesheets/common/base/topic.scss",
+  "line": 34,
+  "why": "Same as golden 0: float: left kept inside the new flexbox header layout (.extra-info-wrapper .badge-wrapper); SCSS, no methods."
+ },
+ {
+  "repo": "discourse",
+  "pr": 6,
+  "bug": "String mutation via the << operator in URL display",
+  "methods": [
+   "app/serializers/user_serializer.rb::UserSerializer.website_name"
+  ],
+  "file": "app/serializers/user_serializer.rb",
+  "line": 149,
+  "why": "website_name uses \".\" << website_host, which mutates the string literal in place instead of concatenating with + or interpolation."
+ },
+ {
+  "repo": "discourse",
+  "pr": 7,
+  "bug": "scale-color $lightness should use $secondary for dark themes",
+  "methods": [],
+  "file": "app/assets/stylesheets/desktop/topic-post.scss",
+  "line": 291,
+  "why": "The greptile line restates the PR title rather than a defect; the most defensible bug is the light-theme value inverted to 70% at .topic-meta-data h5 a (golden 0/3). SCSS, no methods."
+ },
+ {
+  "repo": "discourse",
+  "pr": 8,
+  "bug": "Race conditions in async group member loading",
+  "methods": [
+   "app/assets/javascripts/admin/routes/admin_group_route.js::setupController"
+  ],
+  "file": "app/assets/javascripts/admin/routes/admin_group_route.js",
+  "line": 13,
+  "why": "Same as golden 0: findMembers is fired asynchronously from setupController with no promise handling."
+ },
+ {
+  "repo": "discourse",
+  "pr": 9,
+  "bug": "Thread-safety issue with lazy @loaded_locales",
+  "methods": [
+   "lib/freedom_patches/translate_accelerator.rb::I18n.ensure_loaded!"
+  ],
+  "file": "lib/freedom_patches/translate_accelerator.rb",
+  "line": 63,
+  "why": "Same as golden 0: @loaded_locales ||= [] in ensure_loaded! is not thread-safe."
+ },
+ {
+  "repo": "discourse",
+  "pr": 10,
+  "bug": "NoMethodError in before_validation of EmbeddableHost",
+  "methods": [
+   "app/models/embeddable_host.rb::<top-level>"
+  ],
+  "file": "app/models/embeddable_host.rb",
+  "line": 6,
+  "why": "Same as golden 0: host.sub! in before_validation raises NoMethodError when host is nil."
+ }
+]

greptile/manifest.json

@@ -0,0 +1,1367 @@
+[
+ {
+  "repo": "sentry",
+  "pr": 1,
+  "title": "Enhanced Pagination Performance for High-Volume Audit Logs",
+  "base": "a5d290951def84afdcc4c88d2f1f20023fc36e2a",
+  "head": "8ab88145113dd23a930e23b9cbbcf8b30e4c0b17",
+  "golden_url": "https://github.com/ai-code-review-evaluation/sentry-greptile/pull/1",
+  "comments": [
+   {
+    "comment": "The OptimizedCursorPaginator in organization_auditlogs.py attempts to use negative slice offsets on Django querysets, but Django querysets do not support negative slicing and will raise an AssertionError.",
+    "severity": "High",
+    "category": "bug"
+   },
+   {
+    "comment": "When requests are authenticated with API keys or org auth tokens (which have user_id=None), organization_context.member is None. Line 71 attempts to access organization_context.member.has_global_access without checking if member is None, causing an AttributeError crash when optimized_pagination=true is used, even though the request passed all permission checks with valid org:write scope.",
+    "severity": "High",
+    "category": "bug"
+   },
+   {
+    "comment": "get_item_key assumes a numeric key, but the paginator is used with order_by=-datetime in the audit logs endpoint; calling math.floor/ceil on a datetime will raise a TypeError.",
+    "severity": "High",
+    "category": "bug"
+   },
+   {
+    "comment": "Importing non-existent OptimizedCursorPaginator",
+    "severity": "Low",
+    "category": "bug"
+   }
+  ]
+ },
+ {
+  "repo": "sentry",
+  "pr": 2,
+  "title": "Optimize spans buffer insertion with eviction during insert",
+  "base": "28e3db2520d4ea28c57b08da57b83917ba7b2e15",
+  "head": "8ab88145113dd23a930e23b9cbbcf8b30e4c0b17",
+  "golden_url": "https://github.com/ai-code-review-evaluation/sentry-greptile/pull/2",
+  "comments": [
+   {
+    "comment": "OptimizedCursorPaginator's negative-offset branch in get_result() passes cursor.offset directly as start_offset to queryset slicing when enable_advanced_features is True and cursor.offset < 0.",
+    "severity": "Critical",
+    "category": "bug"
+   },
+   {
+    "comment": "In BasePaginator.get_result(), when cursor.is_prev is True, it sets start_offset = offset without clamping, allowing negative offset values to be passed to queryset slicing.",
+    "severity": "High",
+    "category": "bug"
+   },
+   {
+    "comment": "OptimizedCursorPaginator.get_item_key uses floor/ceil on a datetime key (order_by='-datetime'), causing TypeError.",
+    "severity": "High",
+    "category": "bug"
+   },
+   {
+    "comment": "Accessing .has_global_access on organization_context.member can raise an AttributeError when organization_context is None, particularly for non-superuser non-member users.",
+    "severity": "High",
+    "category": "bug"
+   }
+  ]
+ },
+ {
+  "repo": "sentry",
+  "pr": 3,
+  "title": "feat(upsampling) - Support upsampled error count with performance optimizations",
+  "base": "a5d290951def84afdcc4c88d2f1f20023fc36e2a",
+  "head": "6ad6fe3899f630ccf922ce7ac681147dc7987435",
+  "golden_url": "https://github.com/ai-code-review-evaluation/sentry-greptile/pull/3",
+  "comments": [
+   {
+    "comment": "In _set_sample_rate_from_error_sampling(), the check `if client_sample_rate:` treats 0.0 as falsy, so a sample_rate of 0.0 is silently skipped and never set on normalized_data.",
+    "severity": "Low",
+    "category": "bug"
+   },
+   {
+    "comment": "Using Python’s built-in hash() to build cache keys is non-deterministic across processes (hash randomization), so keys won’t match across workers and invalidate_upsampling_cache may fail to delete them. Use a deterministic serialization of project_ids for the cache key.",
+    "severity": "Low",
+    "category": "data"
+   },
+   {
+    "comment": "The upsampling eligibility check passes the outer dataset instead of the actual dataset used by scoped_dataset. In paths where the query ultimately runs against discover (e.g., dashboard split) while the original dataset is metrics, upsampling may be skipped even when all projects are allowlisted.",
+    "severity": "Medium",
+    "category": "bug"
+   }
+  ]
+ },
+ {
+  "repo": "sentry",
+  "pr": 4,
+  "title": "GitHub OAuth Security Enhancement",
+  "base": "ccd14e450fcefa5b286c06221775f50a49059f6d",
+  "head": "843d7c71926b03f0c5847a0740ac3f7fb3c5fbad",
+  "golden_url": "https://github.com/getsentry/sentry/pull/67876",
+  "comments": [
+   {
+    "comment": "Null reference if github_authenticated_user state is missing",
+    "severity": "Medium",
+    "category": "bug"
+   },
+   {
+    "comment": "OAuth state uses pipeline.signature (static) instead of a per-request random value",
+    "severity": "Medium",
+    "category": "security"
+   },
+   {
+    "comment": "The code attempts to access integration.metadata[sender][login] without checking for the existence of the sender key. This causes a KeyError for integrations where the sender metadata was not set during creation",
+    "severity": "High",
+    "category": "bug"
+   }
+  ]
+ },
+ {
+  "repo": "sentry",
+  "pr": 5,
+  "title": "Replays Self-Serve Bulk Delete System",
+  "base": "49a275847631e231672d01b5dcd66558c069d25b",
+  "head": "ea188e2d736fd6ed27c1dba8aae63b7268e2a7a9",
+  "golden_url": "https://github.com/ai-code-review-evaluation/sentry-greptile/pull/5",
+  "comments": [
+   {
+    "comment": "The BrowserReportingCollectorEndpoint returns HttpResponse objects but the endpoint inherits from Endpoint which expects rest_framework.response.Response objects.",
+    "severity": "Medium",
+    "category": "api"
+   },
+   {
+    "comment": "In BaseDetectorTypeValidator.update(), the validator incorrectly updates instance.type using validated_data.get('detector_type') instead of validated_data.get('type').",
+    "severity": "Medium",
+    "category": "bug"
+   },
+   {
+    "comment": "Using zip(error_ids, events.values()) assumes the get_multi result preserves the input order; dict value order is not guaranteed to match error_ids, so event data can be paired with the wrong ID (missing nodes also shift alignment).",
+    "severity": "Low",
+    "category": "bug"
+   },
+   {
+    "comment": "Hardcoded empty data passed to TableWidgetVisualization causes it to always render an empty table instead of displaying actual results when the feature flag is enabled.",
+    "severity": "Medium",
+    "category": "bug"
+   },
+   {
+    "comment": "Validation logic incorrectly uses .get('timestamp'), which treats a timestamp of 0 as absent. It should directly check 'timestamp' in self.initial_data to ensure proper field presence detection, preventing both 'age=0' and 'timestamp' from passing mutual-exclusion validation.",
+    "severity": "High",
+    "category": "bug"
+   },
+   {
+    "comment": "Analytics events are recorded before the feature flag check, resulting in analytics being logged for all requests, including those that return 404 due to the feature being disabled. This can lead to misleading analytics data.",
+    "severity": "Medium",
+    "category": "bug"
+   }
+  ]
+ },
+ {
+  "repo": "sentry",
+  "pr": 6,
+  "title": "Span Buffer Multiprocess Enhancement with Health Monitoring",
+  "base": "c2c41b0db17dafbcf901412623e874b48ae37a6e",
+  "head": "9bf83b05423beff6dcc542f93e903c562e97469b",
+  "golden_url": "https://github.com/getsentry/sentry/pull/93824",
+  "comments": [
+   {
+    "comment": "In the SpanFlusher flush loop, metric tags inconsistently use 'shard' (e.g., in spans.buffer.flusher.produce and spans.buffer.segment_size_bytes) and 'shards' (in spans.buffer.flusher.wait_produce) for the same shard identifier value. This prevents correct metric aggregation across flush stages — dashboards and alerts filtering by tag 'shard' will miss data tagged as 'shards' and vice versa.",
+    "severity": "Medium",
+    "category": "bug"
+   },
+   {
+    "comment": "The test_basic function in test_consumer.py uses a monkeypatched time.sleep that returns immediately, but then calls time.sleep(0.1) expecting it to actually wait.",
+    "severity": "Low",
+    "category": "test_gap"
+   },
+   {
+    "comment": "Because flusher processes are created via multiprocessing.get_context('spawn').Process, they are instances of multiprocessing.context.SpawnProcess, which on POSIX is not a subclass of multiprocessing.Process, so this isinstance check will always be false and hung processes won't be killed here.",
+    "severity": "High",
+    "category": "bug"
+   },
+   {
+    "comment": "Sleep in test_consumer.py won’t actually wait because time.sleep was monkeypatched above; consider restoring sleep or using a different sync to ensure the flusher has time to process.",
+    "severity": "Medium",
+    "category": "test_gap"
+   },
+   {
+    "comment": "Breaking out of the loop when the deadline has elapsed can skip terminating remaining flusher processes, potentially leaving them running after shutdown; consider ensuring termination is attempted even if the deadline is exceeded.",
+    "severity": "Medium",
+    "category": "concurrency"
+   }
+  ]
+ },
+ {
+  "repo": "sentry",
+  "pr": 7,
+  "title": "feat(ecosystem): Implement cross-system issue synchronization",
+  "base": "19c92a4345beecc750245775de2b3ab9f4f59473",
+  "head": "f9919d789416dd32549d7bb9e238d5df55d5265a",
+  "golden_url": "https://github.com/getsentry/sentry/pull/77754",
+  "comments": [
+   {
+    "comment": "The AssignmentSource dataclass uses queued: datetime = timezone.now() as a default field value, which evaluates timezone.now() once at class definition time and shares that single datetime instance across all objects.",
+    "severity": "Medium",
+    "category": "concurrency"
+   },
+   {
+    "comment": "The method name has a typo: test_from_dict_inalid_data should be test_from_dict_invalid_data.",
+    "severity": "Low",
+    "category": "style"
+   },
+   {
+    "comment": "Method name says 'empty_array' but tests empty dict - consider renaming to 'test_from_dict_empty_dict' for clarity.",
+    "severity": "Low",
+    "category": "style"
+   },
+   {
+    "comment": "to_dict() returns a datetime for queued; if this dict is passed in task kwargs (e.g., via apply_async), JSON serialization may fail depending on the serializer, which can cause enqueue errors.",
+    "severity": "Medium",
+    "category": "bug"
+   }
+  ]
+ },
+ {
+  "repo": "sentry",
+  "pr": 8,
+  "title": "ref(crons): Reorganize incident creation / issue occurrence logic",
+  "base": "5fc9f67c1bd6a50d6c6eb3270b1e6fb7938c5865",
+  "head": "c3b32a7babfcf62b91553388bcd9e9fbfd30a690",
+  "golden_url": "https://github.com/getsentry/sentry/pull/80528",
+  "comments": [
+   {
+    "comment": "The function modifies the config variable to include display values but then returns the original monitor.config instead of the modified version.",
+    "severity": "High",
+    "category": "bug"
+   },
+   {
+    "comment": "The code fetches MonitorCheckIn objects by ID when the required data already exists in previous_checkins. This creates an unnecessary database query.",
+    "severity": "Low",
+    "category": "perf"
+   }
+  ]
+ },
+ {
+  "repo": "sentry",
+  "pr": 9,
+  "title": "feat(uptime): Add ability to use queues to manage parallelism",
+  "base": "5c94383dd49b567fb086ad594b40af066f426858",
+  "head": "33f4fddf64465f75308957c4431e56fd0443eca3",
+  "golden_url": "https://github.com/getsentry/sentry/pull/95633",
+  "comments": [
+   {
+    "comment": "The magic number 50 for max_wait is used repeatedly throughout the tests. Consider extracting this as a named constant to improve maintainability.",
+    "severity": "Low",
+    "category": "style"
+   },
+   {
+    "comment": "The test test_thread_queue_parallel_error_handling has a docstring that doesn't match the test implementation.",
+    "severity": "Low",
+    "category": "doc_defect"
+   }
+  ]
+ },
+ {
+  "repo": "sentry",
+  "pr": 10,
+  "title": "feat(workflow_engine): Add in hook for producing occurrences from the stateful detector",
+  "base": "7249010ca8fea2bbc9ae58e7352861c2d17c2ef2",
+  "head": "de60b7fb01a1df4ecca82cc842a8b5fb16d92794",
+  "golden_url": "https://github.com/getsentry/sentry/pull/80168",
+  "comments": [
+   {
+    "comment": "MetricAlertDetectorHandler inherits from StatefulDetectorHandler but only contains pass, failing to implement its required abstract methods: counter_names (property), get_dedupe_value(), get_group_key_values(), and build_occurrence_and_event_data(). This will cause a TypeError at runtime when the class is instantiated.",
+    "severity": "High",
+    "category": "api"
+   },
+   {
+    "comment": "Docstring says this returns a list of DetectorEvaluationResult, but the method now returns a dict keyed by DetectorGroupKey. Consider updating the docstring to match the new return type.",
+    "severity": "Low",
+    "category": "doc_defect"
+   },
+   {
+    "comment": "Wrong value `6` passed for `group_2` occurrence — the data packet specifies `\"group_2\": 10`. Should be `10` to accurately reflect the test scenario, which will ensure the test produces the correct occurrence data.",
+    "severity": "Medium",
+    "category": "test_gap"
+   }
+  ]
+ },
+ {
+  "repo": "grafana",
+  "pr": 1,
+  "title": "Anonymous: Add configurable device limit",
+  "base": "ed865831071a0c0a63c75ad0cfa6661107dabcaf",
+  "head": "3647ba7360b8d157e4a294a650a365b6aca070d8",
+  "golden_url": "https://github.com/grafana/grafana/pull/79265",
+  "comments": [
+   {
+    "comment": "Race condition: Multiple concurrent requests could pass the device count check simultaneously and create devices beyond the limit. Consider using a database transaction or lock.",
+    "severity": "High",
+    "category": "concurrency"
+   },
+   {
+    "comment": "Anonymous authentication now fails entirely if anonDeviceService.TagDevice returns ErrDeviceLimitReached. Previously, device tagging was asynchronous and non-blocking. This change prevents anonymous users from authenticating when the device limit is reached.",
+    "severity": "Medium",
+    "category": "bug"
+   },
+   {
+    "comment": "The updateDevice function returns ErrDeviceLimitReached when rowsAffected == 0, but zero affected rows means the device doesn't exist or falls outside the time window, not that a limit was reached.",
+    "severity": "Low",
+    "category": "doc_defect"
+   },
+   {
+    "comment": "Time window calculation inconsistency: Using device.UpdatedAt.UTC().Add(-anonymousDeviceExpiration) as the lower bound but device.UpdatedAt as the current time may not match the intended logic. Consider using time.Now().UTC() consistently.",
+    "severity": "Low",
+    "category": "bug"
+   },
+   {
+    "comment": "The `tagDeviceUI` function populates the local cache with the device ID before the `CreateOrUpdateDevice` call. If `ErrDeviceLimitReached` is returned, the cache entry remains set, allowing subsequent requests to bypass the device limit until the cache expires after 29 minutes.",
+    "severity": "High",
+    "category": "bug"
+   }
+  ]
+ },
+ {
+  "repo": "grafana",
+  "pr": 2,
+  "title": "AuthZService: improve authz caching",
+  "base": "e650fa7b2041c91fbb98ef3432b3969e13bbdac9",
+  "head": "d9dc93c4a6c88ef850dbe526cb3226ae99cafa20",
+  "golden_url": "https://github.com/grafana/grafana/pull/103633",
+  "comments": [
+   {
+    "comment": "The Check operation exhibits asymmetric cache trust logic: cached permission grants are trusted and returned immediately, but cached denials from the same permission cache are ignored, leading to a fresh database lookup. This allows stale cached grants to provide access to revoked resources, posing a security risk. ",
+    "severity": "High",
+    "category": "security"
+   },
+   {
+    "comment": "The test comment says the cached permissions 'allow access', but the map stores false for dashboards:uid:dash1, so checkPermission will still treat this scope as not allowed.",
+    "severity": "Low",
+    "category": "doc_defect"
+   },
+   {
+    "comment": "The cache key construction in `userPermDenialCacheKey` uses simple string concatenation with underscores, which can lead to ambiguous keys and collision risks when any component contains underscores or is empty. This could result in incorrect denial cache hits, affecting user permissions.",
+    "severity": "Medium",
+    "category": "data"
+   }
+  ]
+ },
+ {
+  "repo": "grafana",
+  "pr": 3,
+  "title": "Plugins: Chore: Renamed instrumentation middleware to metrics middleware",
+  "base": "5dac6731cbb99f3374898cf6a1558edb222d323b",
+  "head": "00393c2ac9103072bc2194118e455f9248a37816",
+  "golden_url": "https://github.com/grafana/grafana/pull/76186",
+  "comments": [
+   {
+    "comment": "The ContextualLoggerMiddleware methods (QueryData, CallResource, CheckHealth, CollectMetrics) panic when a nil request is received. This occurs because they directly access req.PluginContext (via the instrumentContext function) without first checking if req is nil. This is a regression, as previous middleware layers gracefully handled nil requests.",
+    "severity": "High",
+    "category": "bug"
+   },
+   {
+    "comment": "The traceID is no longer logged for plugin requests. During a refactoring, the tracing import and the logic to extract and add traceID from the context to log parameters were removed from the LoggerMiddleware. The newly introduced ContextualLoggerMiddleware does not add this information, resulting in missing traceID in plugin request logs and impacting debugging and request tracing capabilities.",
+    "severity": "Low",
+    "category": "bug"
+   },
+   {
+    "comment": "The `TestLogger.FromContext` method returns a new `TestLogger` instance via `NewTestLogger()` instead of returning the receiver (f). This causes logs to be recorded on an un-referenced instance, leading to broken test assertions that expect logs to be recorded on the original logger.",
+    "severity": "High",
+    "category": "bug"
+   }
+  ]
+ },
+ {
+  "repo": "grafana",
+  "pr": 4,
+  "title": "Advanced Query Processing Architecture",
+  "base": "dcb853d3099c08b74370fcc965342f035dd6d6cc",
+  "head": "3ba2457a7090b2b76f4b368c8a9db835d0bec66c",
+  "golden_url": "https://github.com/grafana/grafana/pull/107534",
+  "comments": [
+   {
+    "comment": "The applyTemplateVariables method is called with request.filters as the third parameter, but this parameter is not used in the corresponding test setup.",
+    "severity": "Low",
+    "category": "test_gap"
+   }
+  ]
+ },
+ {
+  "repo": "grafana",
+  "pr": 5,
+  "title": "Notification Rule Processing Engine",
+  "base": "acdb0e151c9a2abd6e691353de6cc515d316ee76",
+  "head": "3e6d620d2c1d1f6c10611fba8b0ee062d6fb2a68",
+  "golden_url": "https://github.com/grafana/grafana/pull/106778",
+  "comments": [
+   {
+    "comment": "The rendered GrafanaRuleListItem is missing the required key prop for React list items. This can cause rendering issues when the list order changes.",
+    "severity": "Medium",
+    "category": "bug"
+   },
+   {
+    "comment": "RuleActionsButtons is invoked with only promRule, but SilenceGrafanaRuleDrawer inside RuleActionsButtons still depends on a Grafana Ruler rule being present, so for Grafana rules coming from list views the 'Silence notifications' menu entry (now driven by Grafana Prom abilities) will toggle showSilenceDrawer without ever rendering the drawer. This means clicking 'Silence notifications' for these rules has no visible effect, even when abilities indicate silencing is allowed.",
+    "severity": "High",
+    "category": "bug"
+   }
+  ]
+ },
+ {
+  "repo": "grafana",
+  "pr": 6,
+  "title": "Dual Storage Architecture",
+  "base": "5e3a5b355ef7478d7be86923749e25aebef5226a",
+  "head": "67b74e1e8a7d26d24617b7313b90fa4649138318",
+  "golden_url": "https://github.com/grafana/grafana/pull/90045",
+  "comments": [
+   {
+    "comment": "The context is being created with d.Log instead of the log variable that was initialized with additional context values (name, kind, method). This means those values won't be propagated to the logging context.",
+    "severity": "Medium",
+    "category": "bug"
+   },
+   {
+    "comment": "In DualWriterMode3's DeleteCollection method, when a storage operation fails, recordStorageDuration is incorrectly called with recordLegacyDuration instead.",
+    "severity": "High",
+    "category": "bug"
+   },
+   {
+    "comment": "Inconsistency: using name instead of options.Kind for metrics recording differs from other methods.",
+    "severity": "Medium",
+    "category": "api"
+   }
+  ]
+ },
+ {
+  "repo": "grafana",
+  "pr": 7,
+  "title": "Database Performance Optimizations",
+  "base": "f84c8f685310cc41b52831c130e700a01b225b58",
+  "head": "81c45bfe449584dc7ff50e4fabd5d0c6568999b5",
+  "golden_url": "https://github.com/grafana/grafana/pull/80329",
+  "comments": [
+   {
+    "comment": "The code uses Error log level for what appears to be debugging information. This will pollute error logs in production. Consider using Debug or Info level instead.",
+    "severity": "Low",
+    "category": "perf"
+   },
+   {
+    "comment": "The cleanup job ticker was changed from 10 minutes to 1 minute, which can lead to increased database load and contention due to more frequent execution of cleanup tasks. This change should be reviewed to ensure it aligns with performance expectations.",
+    "severity": "Low",
+    "category": "perf"
+   }
+  ]
+ },
+ {
+  "repo": "grafana",
+  "pr": 8,
+  "title": "Frontend Asset Optimization",
+  "base": "0a870e6a8855cee9c8ae391767da40dbb1c16cc6",
+  "head": "95000f9fc8049d47f837b6e8603c230c2a073ae5",
+  "golden_url": "https://github.com/grafana/grafana/pull/90939",
+  "comments": [
+   {
+    "comment": "The GetWebAssets function implements an incomplete double-checked locking pattern for caching web assets. The function first checks if the cache is populated using a read lock (RLock), and if the cache is empty, it acquires a write lock to populate it. However, it fails to re-check whether the cache was populated by another goroutine while waiting to acquire the write lock.",
+    "severity": "Medium",
+    "category": "concurrency"
+   },
+   {
+    "comment": "In addition to the missing double-check, the function has a critical flaw in its error handling: it unconditionally assigns the fetch result to the cache (line 69: entryPointAssetsCache = result) regardless of whether the fetch succeeded or failed. When an error occurs during asset fetching, result is nil, and this nil value overwrites any previously valid cache entry.",
+    "severity": "High",
+    "category": "bug"
+   }
+  ]
+ },
+ {
+  "repo": "grafana",
+  "pr": 9,
+  "title": "Advanced SQL Analytics Framework",
+  "base": "cbe1e7d63f098e306058c0fbcab2f5c30602fa7d",
+  "head": "ea71201ddc66f06a3f16387e85a09c53514aaadd",
+  "golden_url": "https://github.com/grafana/grafana/pull/94942",
+  "comments": [
+   {
+    "comment": "The enableSqlExpressions function has flawed logic that always returns false, effectively disabling SQL expressions unconditionally:",
+    "severity": "Critical",
+    "category": "bug"
+   },
+   {
+    "comment": "The DB methods NewInMemoryDB().RunCommands, DB.QueryFramesInto, and DB.TablesList in pkg/expr/sql/db.go return hardcoded 'not implemented' errors after removing the go-duck dependency.",
+    "severity": "High",
+    "category": "api"
+   }
+  ]
+ },
+ {
+  "repo": "grafana",
+  "pr": 10,
+  "title": "Unified Storage Performance Optimizations",
+  "base": "871af07203177ce59a24e742107119409274f48a",
+  "head": "d762a9643618a6ba7c43da27374bd9ffc80889ce",
+  "golden_url": "https://github.com/grafana/grafana/pull/97529",
+  "comments": [
+   {
+    "comment": "A race condition in BuildIndex allows multiple goroutines to concurrently build the same expensive index for the same key. This is caused by moving the b.cacheMu lock from protecting the entire function to only protecting the final cache assignment. ",
+    "severity": "High",
+    "category": "concurrency"
+   },
+   {
+    "comment": "Calling s.search.TotalDocs() here may race with concurrent index creation: TotalDocs iterates b.cache without synchronization, and the event watcher goroutine started just above could trigger BuildIndex writes concurrently, potentially causing a concurrent map read/write panic.",
+    "severity": "High",
+    "category": "concurrency"
+   }
+  ]
+ },
+ {
+  "repo": "cal.com",
+  "pr": 2,
+  "title": "Async import of the appStore packages",
+  "base": "bff28e5d6fef42b4e7c036c688f0bad4410b42a2",
+  "head": "e043a5efd872362a3fc146ed0d9ce4393e316c1e",
+  "golden_url": "https://github.com/calcom/cal.com/pull/8087",
+  "comments": [
+   {
+    "comment": "Consider adding try-catch around the await to handle import failures gracefully",
+    "severity": "Low",
+    "category": "speculative"
+   },
+   {
+    "comment": "The code uses forEach with async callbacks, which causes asynchronous operations (e.g., calendar/video event deletions, payment refunds) to run concurrently without being awaited. This 'fire-and-forget' behavior leads to unhandled promise rejections, race conditions, and incomplete cleanup, as surrounding try-catch blocks cannot properly handle errors from these unawaited promises. Replace forEach with for...of loops or Promise.all() with map() to ensure proper sequential execution and error handling.",
+    "severity": "Critical",
+    "category": "concurrency"
+   }
+  ]
+ },
+ {
+  "repo": "cal.com",
+  "pr": 3,
+  "title": "feat: 2fa backup codes",
+  "base": "efa6d464a38e60ceeeb88f40668c1c4ac4bfaf54",
+  "head": "a308075bc39b77ed7059b0cae9d443d669a7bf98",
+  "golden_url": "https://github.com/calcom/cal.com/pull/10600",
+  "comments": [
+   {
+    "comment": "The exported function TwoFactor handles backup codes and is in BackupCode.tsx. Inconsistent naming.",
+    "severity": "Low",
+    "category": "style"
+   },
+   {
+    "comment": "Error message mentions 'backup code login' but this is a disable endpoint, not login",
+    "severity": "Low",
+    "category": "doc_defect"
+   },
+   {
+    "comment": "Backup code validation is case-sensitive due to the use of indexOf(). This causes validation to fail if a user enters uppercase hex characters, as backup codes should be case-insensitive for a better user experience.",
+    "severity": "Medium",
+    "category": "bug"
+   },
+   {
+    "comment": "Because backupCodes are decrypted and mutated in memory before being written back, two concurrent login requests using the same backupCode could both pass this check and update, so a single backup code may effectively be accepted more than once if used concurrently, weakening the intended one-time-use semantics.",
+    "severity": "High",
+    "category": "concurrency"
+   },
+   {
+    "comment": "Object URL for backup codes is not revoked on component unmount, leading to a potential memory leak. This can cause increased memory usage over time, especially if the component is mounted and unmounted frequently.",
+    "severity": "Low",
+    "category": "perf"
+   }
+  ]
+ },
+ {
+  "repo": "cal.com",
+  "pr": 4,
+  "title": "fix: handle collective multiple host on destinationCalendar",
+  "base": "500bfb4d82c53c42e7f53250c6715f5fb474dc2b",
+  "head": "bd797c1ea00a61425735f2b9fd7ce212ebf8c8ab",
+  "golden_url": "https://github.com/calcom/cal.com/pull/10967",
+  "comments": [
+   {
+    "comment": "A null reference occurs in GoogleCalendarService.createEvent() and updateEvent() when mainHostDestinationCalendar is destructured from evt.destinationCalendar but evt.destinationCalendar is null or an empty array, causing mainHostDestinationCalendar to be undefined.",
+    "severity": "High",
+    "category": "bug"
+   },
+   {
+    "comment": "The optional chaining on mainHostDestinationCalendar?.integration is redundant since you already check mainHostDestinationCalendar in the ternary condition.",
+    "severity": "Low",
+    "category": "style"
+   },
+   {
+    "comment": "Logic error: when externalCalendarId is provided, you're searching for a calendar where externalId === externalCalendarId, but this will always fail since you're looking for a calendar that matches itself. Should likely find by credentialId or use different logic.",
+    "severity": "High",
+    "category": "bug"
+   },
+   {
+    "comment": "Logic inversion in organization creation: The slug property is now conditionally set when IS_TEAM_BILLING_ENABLED is true, instead of when it's false as originally intended. This change, combined with requestedSlug still being set when IS_TEAM_BILLING_ENABLED is true, results in both properties being set when billing is enabled, and neither when disabled",
+    "severity": "Medium",
+    "category": "bug"
+   },
+   {
+    "comment": "The Calendar interface now requires createEvent(event, credentialId), but some implementations (e.g., Lark/Office365) still declare createEvent(event) only—this breaks the interface contract (also applies to other locations in the PR).",
+    "severity": "Low",
+    "category": "api"
+   },
+   {
+    "comment": "The `credential` may be undefined when passed to `updateEvent` if `reference.credentialId` is set but the credential is not found. This can lead to a runtime error when `updateEvent` is called with an undefined credential, which should be handled to prevent crashes.",
+    "severity": "High",
+    "category": "bug"
+   }
+  ]
+ },
+ {
+  "repo": "cal.com",
+  "pr": 5,
+  "title": "feat: convert InsightsBookingService to use Prisma.sql raw queries",
+  "base": "202d248ae45f8f2f01324b0d8eb5914e9154112f",
+  "head": "56ff75f07b92ba8ef499ff28a2f98c25f339fdae",
+  "golden_url": "https://github.com/calcom/cal.com/pull/22345",
+  "comments": [
+   {
+    "comment": "In getBaseConditions(), the else if (filterConditions) and final else branches are unreachable. This is because getAuthorizationConditions() always returns a non-null Prisma.Sql object, making authConditions always truthy, which means only the first two if/else if conditions are ever evaluated.",
+    "severity": "Low",
+    "category": "bug"
+   },
+   {
+    "comment": "Fetching userIdsFromOrg only when teamsFromOrg.length > 0 can exclude org-level members for orgs without child teams; consider deriving from teamIds (which includes orgId) or removing the guard so org-only orgs still include member user bookings.",
+    "severity": "Medium",
+    "category": "speculative"
+   }
+  ]
+ },
+ {
+  "repo": "cal.com",
+  "pr": 6,
+  "title": "Comprehensive workflow reminder management for booking lifecycle events",
+  "base": "235ec0e906a189ceb0843cfa7af585b1d46f699f",
+  "head": "2ebfcf448c7e4136ac1b8b9ef9b526a4424fd155",
+  "golden_url": "https://github.com/calcom/cal.com/pull/7232",
+  "comments": [
+   {
+    "comment": "Asynchronous functions deleteScheduledEmailReminder and deleteScheduledSMSReminder are called without await inside forEach loops. This occurs during booking rescheduling/cancellation, and workflow/workflow step deletion/updates. Consequently, scheduled workflow reminders may not be reliably cancelled, potentially leaving them active.",
+    "severity": "Medium",
+    "category": "concurrency"
+   },
+   {
+    "comment": "When immediateDelete is true, the deleteScheduledEmailReminder function cancels the SendGrid email but fails to delete the corresponding WorkflowReminder record from the database. This creates orphaned database entries and is inconsistent with the immediateDelete: false path, which marks the record as cancelled. The SendGrid DELETE API call is also omitted in this path.",
+    "severity": "High",
+    "category": "data"
+   },
+   {
+    "comment": "The 'cancelled' field is nullable, which can lead to existing records having null values instead of a default of false. This can cause incorrect behavior in the application when checking the cancellation status of reminders.",
+    "severity": "Medium",
+    "category": "data"
+   }
+  ]
+ },
+ {
+  "repo": "cal.com",
+  "pr": 7,
+  "title": "Advanced date override handling and timezone compatibility improvements",
+  "base": "c4fe69d8265ba632c10f3a90ea05e4a74ff14c85",
+  "head": "3ef328421f60cd096b7676ef7ee7c32b71593049",
+  "golden_url": "https://github.com/calcom/cal.com/pull/8330",
+  "comments": [
+   {
+    "comment": "Incorrect end time calculation using slotStartTime instead of slotEndTime",
+    "severity": "Medium",
+    "category": "bug"
+   },
+   {
+    "comment": "Using === for dayjs object comparison will always return false as it compares object references, not values. Use .isSame() method instead: dayjs(date.start).add(utcOffset, 'minutes').isSame(dayjs(date.end).add(utcOffset, minutes))",
+    "severity": "Medium",
+    "category": "bug"
+   }
+  ]
+ },
+ {
+  "repo": "cal.com",
+  "pr": 8,
+  "title": "OAuth credential sync and app integration enhancements",
+  "base": "bc89fe00ea84d20bedcec782f0701b9711dc8201",
+  "head": "824145b0e245661875240ed5c8745decd2293d1a",
+  "golden_url": "https://github.com/calcom/cal.com/pull/11059",
+  "comments": [
+   {
+    "comment": "The parseRefreshTokenResponse function incorrectly sets refresh_token to the hardcoded string 'refresh_token' when it's missing from the OAuth refresh token response. This invalidates the token, breaking subsequent token refreshes and causing authentication failures.",
+    "severity": "High",
+    "category": "bug"
+   },
+   {
+    "comment": "The Zod schema in parseRefreshTokenResponse uses computed property keys [z.string().toString()] and [z.string().optional().toString()] in minimumTokenResponseSchema, which is invalid Zod syntax and will throw a runtime error.",
+    "severity": "High",
+    "category": "bug"
+   },
+   {
+    "comment": "parseRefreshTokenResponse returns a Zod safeParse result ({ success, data, error }), not the credential key object. Persisting that as key stores the wrapper instead of the token payload; we should store the parsed data or use schema parse.",
+    "severity": "High",
+    "category": "data"
+   },
+   {
+    "comment": "When APP_CREDENTIAL_SHARING_ENABLED and CALCOM_CREDENTIAL_SYNC_ENDPOINT are set, the refreshFunction helper returns the fetch Response, but several callers (for example GoogleCalendarService.refreshAccessToken expecting res.data, and HubspotCalendarService.refreshAccessToken expecting a HubspotToken) assume it returns the integration-specific token object. That mismatch will cause runtime errors in the sync-enabled path unless the return type or those call sites are adjusted.",
+    "severity": "High",
+    "category": "api"
+   },
+   {
+    "comment": "When the sync endpoint path is used, res is a fetch Response and has no .data; res?.data will be undefined and token.access_token will throw at runtime. This relies on a consistent return shape from refreshOAuthTokens, which isn’t guaranteed currently.",
+    "severity": "High",
+    "category": "bug"
+   },
+   {
+    "comment": "The `refreshOAuthTokens` function is passed `credentialId` instead of `userId`, which breaks the credential sync functionality by causing incorrect user lookups. This issue can lead to failures in syncing user credentials, impacting user experience.",
+    "severity": "High",
+    "category": "bug"
+   },
+   {
+    "comment": "The Salesforce integration checks `response.statusText` instead of `response.ok` when handling OAuth token refresh responses. This can lead to unreliable behavior, as a successful response may not be correctly identified, resulting in the use of stale credentials and potential authentication failures.",
+    "severity": "High",
+    "category": "bug"
+   },
+   {
+    "comment": "The `jsforce.Connection` is created using stale `credentialKey.access_token` and `credentialKey.instance_url` after a token refresh, which can lead to authentication failures and incorrect data access. Ensure that the connection is established using the refreshed token values instead.",
+    "severity": "High",
+    "category": "bug"
+   },
+   {
+    "comment": "The `appCredentialWebhookRequestBodySchema.parse` function can throw on invalid payloads, resulting in 500 Internal Server Errors instead of the expected 400 Bad Request responses. This can lead to confusion for clients and hinder proper error handling.",
+    "severity": "Medium",
+    "category": "api"
+   }
+  ]
+ },
+ {
+  "repo": "cal.com",
+  "pr": 9,
+  "title": "SMS workflow reminder retry count tracking",
+  "base": "917c7b0764f4c6bfdca95db2c5f73dbfdbe11757",
+  "head": "2dae0bfb8c72b51aa49260dc1f7f215a6290f3a6",
+  "golden_url": "https://github.com/calcom/cal.com/pull/14943",
+  "comments": [
+   {
+    "comment": "Using retryCount: reminder.retryCount + 1 reads a possibly stale value and can lose increments under concurrency; consider an atomic increment via Prisma (increment: 1) to avoid race conditions (also applies to the similar update in the catch block).",
+    "severity": "High",
+    "category": "concurrency"
+   },
+   {
+    "comment": "The deletion logic in scheduleSMSReminders.ts incorrectly deletes non-SMS workflow reminders (e.g., Email, WhatsApp) that have retryCount > 1. This occurs because the retryCount condition within the OR clause for deletion lacks a method: WorkflowMethods.SMS filter, causing it to apply to all reminder types instead of only SMS reminders, which is the intended scope of this function.",
+    "severity": "High",
+    "category": "bug"
+   }
+  ]
+ },
+ {
+  "repo": "cal.com",
+  "pr": 10,
+  "title": "Add guest management functionality to existing bookings",
+  "base": "b004587262e8221083bafbe9a0c515e7becaa7b3",
+  "head": "baa9045af7711ea0b86390cb0bf0fdce5430228c",
+  "golden_url": "https://github.com/calcom/cal.com/pull/14740",
+  "comments": [
+   {
+    "comment": "The addGuests.handler.ts blacklist check uses blacklistedGuestEmails.includes(guest) after lowercasing the blacklist but not the input guest emails, allowing attackers to bypass BLACKLISTED_GUEST_EMAILS restrictions by submitting emails with uppercase characters.",
+    "severity": "High",
+    "category": "security"
+   },
+   {
+    "comment": "The logic for checking team admin/owner permissions is incorrect. This condition uses AND (&&) which requires both isTeamAdmin AND isTeamOwner to be true, but it should use OR (||) since a user needs to be either an admin OR an owner to have permission.",
+    "severity": "Critical",
+    "category": "bug"
+   },
+   {
+    "comment": "The sendAddGuestsEmails function in email-manager.ts passes the full calendarEvent.attendees array when determining which attendees receive AttendeeScheduledEmail vs AttendeeAddGuestsEmail, but checks if newGuests.includes(attendee.email) where newGuests is the input guests array.",
+    "severity": "Medium",
+    "category": "bug"
+   },
+   {
+    "comment": "uniqueGuests filters out existing attendees and blacklisted emails but does not deduplicate duplicates within the input; createMany can insert duplicate attendee rows if the client submits repeated emails.",
+    "severity": "Medium",
+    "category": "bug"
+   },
+   {
+    "comment": "Starting with an array containing an empty string may cause validation issues. Consider starting with an empty array [] and handling the empty state in the MultiEmail component instead.",
+    "severity": "Low",
+    "category": "bug"
+   },
+   {
+    "comment": "The `sendAddGuestsEmails` function does not check the `disableStandardEmails` flags in the `EventType` metadata, which may result in sending emails to hosts/attendees even when standard emails are disabled for the event type.",
+    "severity": "Medium",
+    "category": "bug"
+   }
+  ]
+ },
+ {
+  "repo": "cal.com",
+  "pr": 11,
+  "title": "feat: add calendar cache status and actions (#22532)",
+  "base": "3784421f025afc3a595fbb4e222916785cc6de0c",
+  "head": "cb7869234729f908648af6796ab4082b569daf39",
+  "golden_url": "https://github.com/calcom/cal.com/pull/22532",
+  "comments": [
+   {
+    "comment": "The updateManyByCredentialId call uses an empty data object, which prevents Prisma's @updatedAt decorator from updating the updatedAt timestamp. This results in inaccurate cache status tracking, as the timestamp isn't updated when the cache is refreshed. To fix this, explicitly set the updatedAt field.",
+    "severity": "Medium",
+    "category": "bug"
+   },
+   {
+    "comment": "logic: macOS-specific sed syntax with empty string after -i flag will fail on Linux systems",
+    "severity": "Low",
+    "category": "bug"
+   },
+   {
+    "comment": "Throwing a generic Error for authentication failures results in a 500 response in tRPC. Instead, it should return a structured TRPCError with an appropriate code, such as FORBIDDEN, to accurately reflect the error type.",
+    "severity": "High",
+    "category": "api"
+   },
+   {
+    "comment": "Date is formatted with hardcoded 'en-US' locale instead of using the user's locale from useLocale, which can lead to incorrect date representation for users in different locales.",
+    "severity": "Medium",
+    "category": "bug"
+   }
+  ]
+ },
+ {
+  "repo": "keycloak",
+  "pr": 1,
+  "title": "Fixing Re-authentication with passkeys",
+  "base": "30f804af450dec523909a52f2a4b97302d27f5cc",
+  "head": "3214b188de808fdb0fff335556b686f1b0a63218",
+  "golden_url": "https://github.com/ai-code-review-evaluation/keycloak-greptile/pull/1",
+  "comments": [
+   {
+    "comment": "In UsernameForm.authenticate(), isConditionalPasskeysEnabled() is called without passing the required UserModel parameter.",
+    "severity": "Medium",
+    "category": "api"
+   },
+   {
+    "comment": "With isConditionalPasskeysEnabled(UserModel user) requiring user != null, authenticate(...) will not call webauthnAuth.fillContextForm(context) on the initial login page where context.getUser() is still null, so conditional passkey UI will not be set up for first-time passkey login. Consider whether this should also be enabled when no user has been selected yet so ID-less passkey authentication on the initial login form continues to work.",
+    "severity": "Medium",
+    "category": "speculative"
+   }
+  ]
+ },
+ {
+  "repo": "keycloak",
+  "pr": 2,
+  "title": "Add caching support for IdentityProviderStorageProvider.getForLogin operations",
+  "base": "5ff9e9147dc08432b3ec7753906467c390b1e6ee",
+  "head": "3e597722a9164d145bc46aa1de5bd714fb6ba357",
+  "golden_url": "https://github.com/keycloak/keycloak/pull/32918",
+  "comments": [
+   {
+    "comment": "InfinispanIdentityProviderStorageProvider.getForLogin() recursively calls session.identityProviders().getById() instead of idpDelegate.getById(), causing the cache lookup to re-enter the same caching layer.",
+    "severity": "Critical",
+    "category": "bug"
+   },
+   {
+    "comment": "Cleanup reference uses incorrect alias - should be 'idp-alias-' + i instead of 'alias'.",
+    "severity": "Medium",
+    "category": "bug"
+   }
+  ]
+ },
+ {
+  "repo": "keycloak",
+  "pr": 3,
+  "title": "Add AuthzClientCryptoProvider for authorization client cryptographic operations",
+  "base": "10aca5552314d6e5c9226f3c6d0107861c38c693",
+  "head": "b95d12a968e8374a93ddcd56f47ab12eb4136ff4",
+  "golden_url": "https://github.com/keycloak/keycloak/pull/33832",
+  "comments": [
+   {
+    "comment": "Returns wrong provider (default keystore instead of BouncyCastle)",
+    "severity": "High",
+    "category": "bug"
+   },
+   {
+    "comment": "Dead code exists where ASN1Encoder instances are created and written to, but their results are immediately discarded. The actual encoding is performed by new ASN1Encoder instances created in the subsequent return statement, rendering the earlier operations useless.",
+    "severity": "Low",
+    "category": "perf"
+   },
+   {
+    "comment": "When length starts at -1 from indefinite-length encoding, the arithmetic 'length = length - bytes.length' will behave unexpectedly, potentially leading to incorrect processing of encoded data.",
+    "severity": "Medium",
+    "category": "bug"
+   },
+   {
+    "comment": "Abstract order() method on CryptoProvider may break backward compatibility with existing implementations",
+    "severity": "Low",
+    "category": "speculative"
+   }
+  ]
+ },
+ {
+  "repo": "keycloak",
+  "pr": 4,
+  "title": "Add rolling-updates feature flag and compatibility framework",
+  "base": "690b0e4bef15654ea9037b0a26b3337e548c6b48",
+  "head": "0f91e67b9025d978059b21b5c41283e5000a509c",
+  "golden_url": "https://github.com/keycloak/keycloak/pull/36882",
+  "comments": [
+   {
+    "comment": "Incorrect method call for exit codes. The picocli.exit() method calls System.exit() directly, which is problematic:",
+    "severity": "Medium",
+    "category": "bug"
+   },
+   {
+    "comment": "Need to add changelog/release notes documenting the exit code change and migration guidance",
+    "severity": "Low",
+    "category": "doc_defect"
+   }
+  ]
+ },
+ {
+  "repo": "keycloak",
+  "pr": 5,
+  "title": "Add Client resource type and scopes to authorization schema",
+  "base": "baa6ebbeb8791017c01ce9fa63965f8dc584ddfe",
+  "head": "7e5540896bbbafc02234b6ea2b15215dc24cb972",
+  "golden_url": "https://github.com/keycloak/keycloak/pull/36880",
+  "comments": [
+   {
+    "comment": "Inconsistent feature flag bug causing orphaned permissions. The AdminPermissions event listener, responsible for cleaning up permissions upon role, client, or group removal, is incorrectly guarded by the ADMIN_FINE_GRAINED_AUTHZ (V1) feature flag. This is inconsistent with other methods in the class that use ADMIN_FINE_GRAINED_AUTHZ_V2. Consequently, if ADMIN_FINE_GRAINED_AUTHZ_V2 is enabled but V1 is not, the permission cleanup logic will not execute, leading to orphaned permission data. Cleanup should occur regardless of which fine-grained authorization version is enabled.",
+    "severity": "High",
+    "category": "bug"
+   },
+   {
+    "comment": "In hasPermission(ClientModel client, String scope), the resource lookup uses findByName(server, client.getId(), server.getId()), but AdminPermissionsSchema.getOrCreateResource creates per-client resources with the owner set to resourceServer.getClientId(), so this lookup will never find those resources and will always fall back to the 'all-clients' resource, effectively ignoring client-specific permissions.",
+    "severity": "High",
+    "category": "bug"
+   },
+   {
+    "comment": "In getClientsWithPermission(String scope), iterating resourceStore.findByType(server, AdminPermissionsSchema.CLIENTS_RESOURCE_TYPE) and returning resource.getName() will only ever consider the type-level 'Clients' resource (per-client resources have no type) and return its name, while AvailableRoleMappingResource#getRoleIdsWithPermissions expects actual client IDs to pass to realm.getClientById, which can lead to incorrect behavior or a null client and subsequent failures.",
+    "severity": "High",
+    "category": "bug"
+   },
+   {
+    "comment": "In ClientPermissionsV2.hasPermission, the method getResourceTypeResource could return null, leading to a NullPointerException in findByResource. Defensive null checks are needed to prevent this issue.",
+    "severity": "High",
+    "category": "bug"
+   },
+   {
+    "comment": "Remove unused getEvaluationContext private method in ClientPermissionsV2 - it is defined but never called",
+    "severity": "Low",
+    "category": "style"
+   }
+  ]
+ },
+ {
+  "repo": "keycloak",
+  "pr": 6,
+  "title": "Add Groups resource type and scopes to authorization schema",
+  "base": "7a8d18122aea42278332120093d5cefce9748356",
+  "head": "679f44692dbc288bedc8f1535071c8467dbaf70c",
+  "golden_url": "https://github.com/keycloak/keycloak/pull/37038",
+  "comments": [
+   {
+    "comment": "In GroupPermissionsV2.canManage(), the permission fallback checks for both VIEW and MANAGE scopes (hasPermission(null, AdminPermissionsSchema.VIEW, AdminPermissionsSchema.MANAGE)), but canManage() should only grant access for MANAGE scope, not VIEW. This means users with only VIEW permission on groups can pass the canManage() check, granting them write access they should not have. The canView() method correctly checks both VIEW and MANAGE, but canManage() should only check MANAGE.",
+    "severity": "High",
+    "category": "bug"
+   },
+   {
+    "comment": "In getGroupIdsWithViewPermission, hasPermission is called with groupResource.getId() and the same groupResource.getId() is added to granted, but hasPermission resolves resources by name (treating the argument as a group id) and the GroupPermissionEvaluator contract says this method returns group IDs that are later used as UserModel.GROUPS and in getUsersCount group filters. This mismatch means per-group VIEW_MEMBERS/MANAGE_MEMBERS permissions may not yield the expected group IDs for filtering and counts, and evaluation may effectively only look at the type-level 'all-groups' resource; consider revisiting whether this should operate on the underlying group ids (resource names) instead so it aligns with the JPA queries and the interface contract.",
+    "severity": "High",
+    "category": "api"
+   }
+  ]
+ },
+ {
+  "repo": "keycloak",
+  "pr": 7,
+  "title": "Add HTML sanitizer for translated message resources",
+  "base": "21d53112849fe55f328f3de2f071989b8763de61",
+  "head": "5d77e7ea7d431aa0a6b65904457b7471bd050e81",
+  "golden_url": "https://github.com/keycloak/keycloak/pull/37429",
+  "comments": [
+   {
+    "comment": "The translation is in Italian instead of Lithuanian. This should be translated to Lithuanian to match the file's locale (messages_lt.properties).",
+    "severity": "Medium",
+    "category": "doc_defect"
+   },
+   {
+    "comment": "The totpStep1 value uses Traditional Chinese terms in the Simplified Chinese file (zh_CN), which is likely incorrect for this locale. Please verify the locale‑appropriate translation.",
+    "severity": "Medium",
+    "category": "doc_defect"
+   },
+   {
+    "comment": "The anchor sanitization logic has a potential issue where it consumes English matcher groups without proper validation. If the translated text has more anchor tags than the English text, this could lead to incorrect validation results.",
+    "severity": "Low",
+    "category": "security"
+   },
+   {
+    "comment": "The method name 'santizeAnchors' should be 'sanitizeAnchors' (missing 'i').",
+    "severity": "Low",
+    "category": "style"
+   },
+   {
+    "comment": "Field HTML_TAGS should be declared as 'private static final' following Java naming conventions for constants",
+    "severity": "Low",
+    "category": "style"
+   }
+  ]
+ },
+ {
+  "repo": "keycloak",
+  "pr": 8,
+  "title": "Implement access token context encoding framework",
+  "base": "b545339e2c924263a498653f348610de0c09059c",
+  "head": "c8738721df2cd2d16fe16adec571b4192354a771",
+  "golden_url": "https://github.com/keycloak/keycloak/pull/37634",
+  "comments": [
+   {
+    "comment": "In AccessTokenContext constructor, the fourth Objects.requireNonNull validates grantType twice instead of validating rawTokenId, allowing null rawTokenId values to pass validation.",
+    "severity": "Critical",
+    "category": "bug"
+   },
+   {
+    "comment": "In isAccessTokenId, the substring for the grant shortcut and the equality check look inverted: the grant shortcut occupies indices 4–5 (substring(4,6)), and a match should return true (combined with UUID check), not false.",
+    "severity": "High",
+    "category": "bug"
+   },
+   {
+    "comment": "Javadoc mentions \"usually like 3-letters shortcut\" but some implementations use 2-letter shortcuts (\"ac\", \"cc\", \"rt\", \"te\", \"pc\", \"ci\", \"ro\"). Consider updating documentation to reflect actual usage pattern.",
+    "severity": "Low",
+    "category": "doc_defect"
+   },
+   {
+    "comment": " Catching generic RuntimeException is too broad. The implementation throws IllegalArgumentException specifically - catch that instead for more precise testing.",
+    "severity": "Low",
+    "category": "bug"
+   }
+  ]
+ },
+ {
+  "repo": "keycloak",
+  "pr": 9,
+  "title": "Implement recovery key support for user storage providers",
+  "base": "a0852eaa2ed19408ecf8ad02652802d00529a085",
+  "head": "134437a5a7e31472763ffa2ac3cc91030e7ee389",
+  "golden_url": "https://github.com/keycloak/keycloak/pull/38446",
+  "comments": [
+   {
+    "comment": "Unsafe raw List deserialization without type safety. Calling Optional.get() directly on the Optional returned by RecoveryAuthnCodesUtils.getCredential(user) without checking isPresent() can lead to a NoSuchElementException if the Optional is empty.",
+    "severity": "Medium",
+    "category": "bug"
+   },
+   {
+    "comment": "After creating the RecoveryAuthnCodesCredentialModel, consider setting its id from the stored credential (e.g., myUser.recoveryCodes.getId()); otherwise getId() will be null and downstream removal by id (e.g., removeStoredCredentialById in the authenticator flow) may not work.",
+    "severity": "Low",
+    "category": "bug"
+   }
+  ]
+ },
+ {
+  "repo": "keycloak",
+  "pr": 10,
+  "title": "Fix concurrent group access to prevent NullPointerException",
+  "base": "e92b825a14800e42f42d5f22ee314e94cf7c4e27",
+  "head": "332c9b6e4a1288705f173a9b797742859384d1e0",
+  "golden_url": "https://github.com/keycloak/keycloak/pull/40940",
+  "comments": [
+   {
+    "comment": "Returning null from getSubGroupsCount() violates the GroupModel contract (Javadoc says it never returns null) and may lead to NPEs in callers that expect a non-null count.",
+    "severity": "Critical",
+    "category": "api"
+   },
+   {
+    "comment": "The reader thread isn’t waited for; flipping deletedAll to true and asserting immediately can race and miss exceptions added just after the flag change, making this test flaky.",
+    "severity": "Medium",
+    "category": "concurrency"
+   }
+  ]
+ },
+ {
+  "repo": "discourse",
+  "pr": 1,
+  "title": "FEATURE: automatically downsize large images",
+  "base": "3f71fa15c93bb0b751bb92abae1b5dc9a21bd4ca",
+  "head": "ffbaf8c54269df2ce510de91245760fddce09896",
+  "golden_url": "https://github.com/ai-code-review-evaluation/discourse-graphite/pull/1",
+  "comments": [
+   {
+    "comment": "The downsize method is defined twice. The second definition, which expects a single dimensions string parameter, overrides the first, which expected separate max_width and max_height parameters. This makes the original method unreachable and breaks existing code that calls it with separate width and height arguments.",
+    "severity": "Medium",
+    "category": "bug"
+   },
+   {
+    "comment": "Hardcoding maxSizeKB = 10 * 1024 ignores Discourse.SiteSettings['max_' + type + '_size_kb'], so the client-side limit can diverge from server-side and per-type settings (also applies to the 413 handler below).",
+    "severity": "Low",
+    "category": "bug"
+   },
+   {
+    "comment": "Passing 80% as the dimensions can fail for animated GIFs when allow_animated_thumbnails is true, since the animated path uses gifsicle --resize-fit which expects WxH geometry, not a percentage; downsizing would then silently fail.",
+    "severity": "Medium",
+    "category": "data"
+   },
+   {
+    "comment": "The loop checking `tempfile.size` may never terminate correctly due to stale cached size, leading to potential infinite loops or incorrect file handling after modifications.",
+    "severity": "Medium",
+    "category": "bug"
+   }
+  ]
+ },
+ {
+  "repo": "discourse",
+  "pr": 2,
+  "title": "FEATURE: per-topic unsubscribe option in emails",
+  "base": "56abd247e1829cb6de82dfcc8ce04deb5f0d1e91",
+  "head": "6669a2d94d76eea3b99b8c476d12b1eb66726b07",
+  "golden_url": "https://github.com/ai-code-review-evaluation/discourse-graphite/pull/2",
+  "comments": [
+   {
+    "comment": "logic: Potential nil pointer exception - if no TopicUser record exists, tu will be nil and calling methods on it will crash",
+    "severity": "High",
+    "category": "bug"
+   },
+   {
+    "comment": "Typo in property name: 'stopNotificiationsText' should be 'stopNotificationsText' (missing 'n' in 'Notifications')",
+    "severity": "Low",
+    "category": "style"
+   }
+  ]
+ },
+ {
+  "repo": "discourse",
+  "pr": 3,
+  "title": "Add comprehensive email validation for blocked users",
+  "base": "e25638dab0d4b98f99c8fe8976ccaae8f4fb9db3",
+  "head": "5f8a130277dbddc95d133cd2832be639baf89213",
+  "golden_url": "https://github.com/ai-code-review-evaluation/discourse-graphite/pull/3",
+  "comments": [
+   {
+    "comment": "BlockedEmail.should_block_email? method has side effects during a read operation - it updates statistics even when just checking if an email should be blocked. This could cause race conditions in concurrent environments and makes the method name misleading.",
+    "severity": "Medium",
+    "category": "concurrency"
+   },
+   {
+    "comment": "The regex pattern @(#{domains}) in email_in_restriction_setting? method only matches domain suffixes, not full domains with proper boundary anchoring. An attacker can register evil.example.com to bypass a whitelist entry for example.com.",
+    "severity": "Medium",
+    "category": "bug"
+   },
+   {
+    "comment": "The case-sensitive email lookup in `BlockedEmail.should_block?` creates a bypass risk, allowing 'User@Example.com' to bypass a block on 'user@example.com'. This can lead to unauthorized access to blocked users.",
+    "severity": "Critical",
+    "category": "security"
+   }
+  ]
+ },
+ {
+  "repo": "discourse",
+  "pr": 4,
+  "title": "Enhance embed URL handling and validation system",
+  "base": "62db063e1e1abe691313ab42682a68f796d63769",
+  "head": "4f8aed295a29954023b2849c060ef4fb299d1b5d",
+  "golden_url": "https://github.com/ai-code-review-evaluation/discourse-graphite/pull/4",
+  "comments": [
+   {
+    "comment": "The PollFeed job calls open(SiteSetting.feed_polling_url) without validating the URL, allowing an attacker who controls the feed_polling_url site setting to make the server fetch arbitrary internal or external URLs.",
+    "severity": "Critical",
+    "category": "security"
+   },
+   {
+    "comment": "The current origin validation using indexOf is insufficient and can be bypassed. An attacker could use a malicious domain like evil-discourseUrl.com to pass this check.",
+    "severity": "Medium",
+    "category": "security"
+   },
+   {
+    "comment": "postMessage targetOrigin should be the origin (scheme+host+port), not the full referrer URL; using the full URL will cause the message to be dropped and prevent resizing.",
+    "severity": "Medium",
+    "category": "security"
+   },
+   {
+    "comment": "The code sets X-Frame-Options: ALLOWALL which completely disables clickjacking protection. The referer validation can be bypassed (referer headers are easily spoofed), and the fallback to empty string for nil referer masks validation failures.",
+    "severity": "Medium",
+    "category": "security"
+   },
+   {
+    "comment": "The TopicEmbed.import method is susceptible to a NoMethodError if the contents parameter is nil when attempting to append a string, and an XSS vulnerability due to unescaped url interpolation in the generated HTML.",
+    "severity": "Medium",
+    "category": "bug"
+   },
+   {
+    "comment": "The ERB block closes with end if, which is invalid Ruby/ERB and will raise at render; it should just be end to close the if block.",
+    "severity": "Medium",
+    "category": "bug"
+   },
+   {
+    "comment": "Calling `i.content.scrub` assumes that `i.content` is always populated. However, many RSS feeds do not provide a `content` field, leading to a potential nil pointer exception when `i.content` is nil. A fallback to `description` should be implemented to prevent crashes.",
+    "severity": "High",
+    "category": "bug"
+   },
+   {
+    "comment": "An unescaped request header (request.referer) is interpolated into JavaScript without proper escaping, which can lead to XSS vulnerabilities if the referer contains malicious content. This issue can allow attackers to execute arbitrary scripts in the context of the application.",
+    "severity": "Critical",
+    "category": "security"
+   }
+  ]
+ },
+ {
+  "repo": "discourse",
+  "pr": 5,
+  "title": "Optimize header layout performance with flexbox mixins",
+  "base": "98fa098e6fade3097a11fe7f99a98a4e6ffb4728",
+  "head": "5b229316ee4c661836ed1161139692a3e8527444",
+  "golden_url": "https://github.com/ai-code-review-evaluation/discourse-graphite/pull/5",
+  "comments": [
+   {
+    "comment": "Mixing float: left with flexbox causes layout issues. Further this PR removes the float-based right alignment for .d-header .panel, which may cause the login panel in the non-Ember/noscript header (where .panel is nested inside .row and not a flex item) to stack under the title instead of remaining right-aligned.",
+    "severity": "Low",
+    "category": "bug"
+   },
+   {
+    "comment": "-ms-align-items never existed in any version of IE/Edge; the correct legacy property is -ms-flex-align.",
+    "severity": "Low",
+    "category": "bug"
+   },
+   {
+    "comment": "-webkit-box-ordinal-group and -moz-box-ordinal-group require values starting from 1, not 0. If $val is 0, this will cause incorrect rendering in older browsers, leading to layout issues.",
+    "severity": "Medium",
+    "category": "bug"
+   }
+  ]
+ },
+ {
+  "repo": "discourse",
+  "pr": 6,
+  "title": "UX: show complete URL path if website domain is same as instance domain",
+  "base": "7a5fbae060bd0d3e29f3a4b5c6f2bfd8b2914dc7",
+  "head": "267d8be1f556ed59639ced396c885bb44586da19",
+  "golden_url": "https://github.com/ai-code-review-evaluation/discourse-graphite/pull/6",
+  "comments": [
+   {
+    "comment": "The include_website_name method is missing the required ? suffix. Rails serializers expect include_ methods to end with ? for conditional attribute inclusion, a convention followed by other methods in this serializer. Without it, the website_name attribute may not be conditionally included as intended. Additionally, the '.' << website_host string concatenation should be replaced with '.' + website_host or '.#{website_host}' to avoid mutating string literals, which can lead to issues.",
+    "severity": "Medium",
+    "category": "api"
+   }
+  ]
+ },
+ {
+  "repo": "discourse",
+  "pr": 7,
+  "title": "scale-color $lightness must use $secondary for dark themes",
+  "base": "7554b5e3c458dc79b591d5e874e79ab10cf30c2f",
+  "head": "d38c4d5f7443223c81529c22470293771baf9f38",
+  "golden_url": "https://github.com/ai-code-review-evaluation/discourse-graphite/pull/7",
+  "comments": [
+   {
+    "comment": "In .topic-meta-data h5 a, the original code had color: scale-color($primary, $lightness: 30%) but was changed to dark-light-choose(scale-color($primary, $lightness: 70%), scale-color($secondary, $lightness: 30%)). The lightness for the light theme changed from 30% to 70%, which is a dramatic inversion",
+    "severity": "Low",
+    "category": "style"
+   },
+   {
+    "comment": "This change for desktop/user.css changes $primary from 30% to 50% for the light theme; most other changes preserve the original $primary value and move the complement to $secondary for dark. Consider reviewing this (also applies to a similar .name change in the mobile variant).",
+    "severity": "Low",
+    "category": "speculative"
+   },
+   {
+    "comment": "In topic-post.css the original code used $lightness: 70% but the replacement uses $lightness: 30% for the light theme. This makes the text significantly darker than intended.",
+    "severity": "Low",
+    "category": "bug"
+   },
+   {
+    "comment": "The light theme color for `.topic-meta-data h5 a` was changed from `scale-color($primary, $lightness: 30%)` to `scale-color($primary, $lightness: 70%)`, which may unintentionally reduce contrast and affect readability. This change should be reviewed to ensure it aligns with design intentions.",
+    "severity": "Medium",
+    "category": "bug"
+   },
+   {
+    "comment": "In mobile/user.scss line 497, the lightness value for dark themes should be inverted to 70% instead of being changed to 50%. This change may lead to unintended visual regressions in light themes.",
+    "severity": "Medium",
+    "category": "bug"
+   }
+  ]
+ },
+ {
+  "repo": "discourse",
+  "pr": 8,
+  "title": "FIX: proper handling of group memberships",
+  "base": "4975fc28903a84418f546b2809370f19abf10e08",
+  "head": "060cda77729cb1c4a827560e09e89a7b22078ba9",
+  "golden_url": "https://github.com/ai-code-review-evaluation/discourse-graphite/pull/8",
+  "comments": [
+   {
+    "comment": " The findMembers() call is now asynchronous and unhandled. The controller may not have member data immediately available, creating a race condition.",
+    "severity": "High",
+    "category": "concurrency"
+   },
+   {
+    "comment": "In the next action, capping the next offset at user_count can produce an empty page (e.g., total equal to limit results in offset == total, showing 2/2 with no members). This can cause confusing UX on the last page.",
+    "severity": "Medium",
+    "category": "bug"
+   },
+   {
+    "comment": "The remove_member test uses xhr :put to call the remove_member action, but the route defines delete 'members' => 'groups#remove_member' requiring DELETE method.",
+    "severity": "Medium",
+    "category": "api"
+   },
+   {
+    "comment": "The `totalPages` calculation using `Math.floor(user_count / limit) + 1` produces an extra page when `user_count` is an exact multiple of `limit`, leading to incorrect pagination. For example, with 100 users and a limit of 50, it reports 3 pages instead of 2, allowing navigation to an empty page.",
+    "severity": "Medium",
+    "category": "bug"
+   },
+   {
+    "comment": "Expected to return a value at the end of function - code paths without explicit returns will return undefined",
+    "severity": "Medium",
+    "category": "bug"
+   },
+   {
+    "comment": "The route provides the group's ID in params[:id], but the code attempts to read it from params[:group_id], causing the action to fail. This results in a broken feature when trying to access group memberships.",
+    "severity": "High",
+    "category": "bug"
+   }
+  ]
+ },
+ {
+  "repo": "discourse",
+  "pr": 9,
+  "title": "FEATURE: Localization fallbacks (server-side)",
+  "base": "728845d008f06828630a632e5c7d67e11bf15deb",
+  "head": "ecfa17b5a79dfdc91e7a4d50b42ae78a35d0a293",
+  "golden_url": "https://github.com/ai-code-review-evaluation/discourse-graphite/pull/9",
+  "comments": [
+   {
+    "comment": "The @loaded_locales instance variable in translate_accelerator.rb is lazily initialized and accessed by multiple concurrent threads without synchronization.",
+    "severity": "Low",
+    "category": "concurrency"
+   },
+   {
+    "comment": "Consider normalizing the input locale (e.g., to a symbol) when checking/loading here to avoid double-loading if the same locale is passed as a String vs Symbol (also applies to other locations in the PR).",
+    "severity": "Low",
+    "category": "perf"
+   }
+  ]
+ },
+ {
+  "repo": "discourse",
+  "pr": 10,
+  "title": "FEATURE: Can edit category/host relationships for embedding",
+  "base": "913c3d6f636d02b3dd6c2c022ca1af2bec95d752",
+  "head": "d1c69189f3c90ecf56013a8da904da9bff9a8e19",
+  "golden_url": "https://github.com/ai-code-review-evaluation/discourse-graphite/pull/10",
+  "comments": [
+   {
+    "comment": "The EmbeddableHost model's before_validation callback calls self.host.sub! to strip http://, https://, and path segments, but if self.host is nil, this raises NoMethodError.",
+    "severity": "Critical",
+    "category": "bug"
+   },
+   {
+    "comment": "The update and destroy methods in Admin::EmbeddableHostsController do not validate the existence of the EmbeddableHost record retrieved by ID. If EmbeddableHost.where(id: params[:id]).first returns nil (i.e., the host does not exist), attempting to call methods on the nil object (e.g., save_host or destroy) will result in a NoMethodError.",
+    "severity": "Medium",
+    "category": "bug"
+   },
+   {
+    "comment": "record_for_host compares lower(host) = ? but does not normalize the parameter’s case, so mixed‑case referer hosts may fail to match even though comparison intends to be case‑insensitive.",
+    "severity": "Medium",
+    "category": "bug"
+   },
+   {
+    "comment": "Because this migration inserts embeddable_hosts rows with raw SQL, any existing embeddable_hosts values that include http:// or /https:// or path segments won’t go through the EmbeddableHost model’s normalization, so the new host lookup (which compares only the bare host) may fail for migrated data. Consider ensuring that migrated hosts are normalized to the same format as newly created EmbeddableHost records so existing embedding configurations keep working.",
+    "severity": "High",
+    "category": "data"
+   },
+   {
+    "comment": "The contents of `category_fabricator.rb` and `embeddable_host_fabricator.rb` have been swapped, resulting in the loss of original category fabricators and potentially causing errors in fabricator usage.",
+    "severity": "High",
+    "category": "bug"
+   },
+   {
+    "comment": "The SQL query in the `app/controllers/admin/embeddable_hosts_controller.rb` file uses string interpolation (`'#{h}'`) to insert values, which is vulnerable to SQL injection. It should use parameter binding or proper quoting to prevent this security vulnerability.",
+    "severity": "Critical",
+    "category": "security"
+   },
+   {
+    "comment": "The migration accesses execute(...)[0]['id'] without checking if the query returned any rows, which causes a crash when no embed_category setting exists. This can lead to a broken feature when attempting to edit category/host relationships.",
+    "severity": "High",
+    "category": "bug"
+   }
+  ]
+ }
+]

greptile/results/jev.jsonl.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:af03c66bc1851b1625a67fb87cc16ba61888a6f32f32e5fc20457fd7ef40d39c
+size 62958

greptile/results/jev048.jsonl.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d4dc953851ffe50c0e6b682842162124932e5da009fc2695bfa17e53e6ab5757
+size 26975

greptile/results/oai.jsonl.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fbbe8bf974f3005181d730dffa6436847e1433b26e130ee43a4524141f625fab
+size 62175

greptile/results/oai2q.jsonl.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2c75ef09d198958e314156a542417d136226a94f934dfeba213de7298d9f040c
+size 26134

greptile/results/oai407.jsonl.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1b0c9c4fd3ea2f1895b0b1572367882808b22e5a8006dedf8613fce3b404e83f
+size 67017

greptile/summary.json

@@ -0,0 +1,302 @@
+[
+ {
+  "config": "oai2q",
+  "label": "perch#418, OpenAI Decisions",
+  "cutoff": 0.6,
+  "prs": 50,
+  "missing": [],
+  "known": 38,
+  "known_of": 50,
+  "known_chance": 20.2,
+  "caught": 42,
+  "of": 47,
+  "chance": 32,
+  "golden": 116,
+  "golden_of": 140,
+  "flags": 490,
+  "read": 1568,
+  "flag_rate": 0.313,
+  "median_flags": 7,
+  "cost": 0.335
+ },
+ {
+  "config": "oai2q",
+  "label": "perch#418, OpenAI Decisions",
+  "cutoff": 0.7,
+  "prs": 50,
+  "missing": [],
+  "known": 33,
+  "known_of": 50,
+  "known_chance": 16.3,
+  "caught": 39,
+  "of": 47,
+  "chance": 27.3,
+  "golden": 105,
+  "golden_of": 140,
+  "flags": 383,
+  "read": 1568,
+  "flag_rate": 0.244,
+  "median_flags": 6,
+  "cost": 0.335
+ },
+ {
+  "config": "oai2q",
+  "label": "perch#418, OpenAI Decisions",
+  "cutoff": 0.8,
+  "prs": 50,
+  "missing": [],
+  "known": 28,
+  "known_of": 50,
+  "known_chance": 11.8,
+  "caught": 39,
+  "of": 47,
+  "chance": 21.5,
+  "golden": 86,
+  "golden_of": 140,
+  "flags": 264,
+  "read": 1568,
+  "flag_rate": 0.168,
+  "median_flags": 4,
+  "cost": 0.335
+ },
+ {
+  "config": "jev048",
+  "label": "perch 0.4.8, Jev",
+  "cutoff": 0.6,
+  "prs": 50,
+  "missing": [],
+  "known": 22,
+  "known_of": 50,
+  "known_chance": 8.2,
+  "caught": 27,
+  "of": 47,
+  "chance": 15.2,
+  "golden": 65,
+  "golden_of": 140,
+  "flags": 179,
+  "read": 1568,
+  "flag_rate": 0.114,
+  "median_flags": 2,
+  "cost": 0.149
+ },
+ {
+  "config": "jev048",
+  "label": "perch 0.4.8, Jev",
+  "cutoff": 0.7,
+  "prs": 50,
+  "missing": [],
+  "known": 12,
+  "known_of": 50,
+  "known_chance": 4,
+  "caught": 19,
+  "of": 47,
+  "chance": 8.8,
+  "golden": 37,
+  "golden_of": 140,
+  "flags": 75,
+  "read": 1568,
+  "flag_rate": 0.048,
+  "median_flags": 0,
+  "cost": 0.149
+ },
+ {
+  "config": "jev048",
+  "label": "perch 0.4.8, Jev",
+  "cutoff": 0.8,
+  "prs": 50,
+  "missing": [],
+  "known": 6,
+  "known_of": 50,
+  "known_chance": 0.7,
+  "caught": 8,
+  "of": 47,
+  "chance": 2.2,
+  "golden": 13,
+  "golden_of": 140,
+  "flags": 18,
+  "read": 1568,
+  "flag_rate": 0.011,
+  "median_flags": 0,
+  "cost": 0.149
+ },
+ {
+  "config": "oai407",
+  "label": "perch#407, OpenAI Decisions",
+  "cutoff": 0.6,
+  "prs": 50,
+  "missing": [],
+  "known": 39,
+  "known_of": 50,
+  "known_chance": 16.8,
+  "caught": 44,
+  "of": 47,
+  "chance": 28.3,
+  "golden": 122,
+  "golden_of": 140,
+  "flags": 918,
+  "read": 4205,
+  "flag_rate": 0.218,
+  "median_flags": 14,
+  "cost": 0.64
+ },
+ {
+  "config": "oai407",
+  "label": "perch#407, OpenAI Decisions",
+  "cutoff": 0.7,
+  "prs": 50,
+  "missing": [],
+  "known": 36,
+  "known_of": 50,
+  "known_chance": 13.1,
+  "caught": 42,
+  "of": 47,
+  "chance": 23.8,
+  "golden": 110,
+  "golden_of": 140,
+  "flags": 684,
+  "read": 4205,
+  "flag_rate": 0.163,
+  "median_flags": 10,
+  "cost": 0.64
+ },
+ {
+  "config": "oai407",
+  "label": "perch#407, OpenAI Decisions",
+  "cutoff": 0.8,
+  "prs": 50,
+  "missing": [],
+  "known": 33,
+  "known_of": 50,
+  "known_chance": 9.4,
+  "caught": 40,
+  "of": 47,
+  "chance": 18.3,
+  "golden": 98,
+  "golden_of": 140,
+  "flags": 472,
+  "read": 4205,
+  "flag_rate": 0.112,
+  "median_flags": 6,
+  "cost": 0.64
+ },
+ {
+  "config": "jev",
+  "label": "perch 0.4.7, Jev",
+  "cutoff": 0.6,
+  "prs": 50,
+  "missing": [],
+  "known": 18,
+  "known_of": 50,
+  "known_chance": 6.1,
+  "caught": 27,
+  "of": 47,
+  "chance": 12.8,
+  "golden": 61,
+  "golden_of": 140,
+  "flags": 418,
+  "read": 4617,
+  "flag_rate": 0.091,
+  "median_flags": 4,
+  "cost": 0.76
+ },
+ {
+  "config": "jev",
+  "label": "perch 0.4.7, Jev",
+  "cutoff": 0.7,
+  "prs": 50,
+  "missing": [],
+  "known": 10,
+  "known_of": 50,
+  "known_chance": 3.2,
+  "caught": 17,
+  "of": 47,
+  "chance": 6.9,
+  "golden": 40,
+  "golden_of": 140,
+  "flags": 176,
+  "read": 4617,
+  "flag_rate": 0.038,
+  "median_flags": 1,
+  "cost": 0.76
+ },
+ {
+  "config": "jev",
+  "label": "perch 0.4.7, Jev",
+  "cutoff": 0.8,
+  "prs": 50,
+  "missing": [],
+  "known": 4,
+  "known_of": 50,
+  "known_chance": 1,
+  "caught": 7,
+  "of": 47,
+  "chance": 2.7,
+  "golden": 13,
+  "golden_of": 140,
+  "flags": 50,
+  "read": 4617,
+  "flag_rate": 0.011,
+  "median_flags": 0,
+  "cost": 0.76
+ },
+ {
+  "config": "oai",
+  "label": "perch 0.4.7, OpenAI Decisions",
+  "cutoff": 0.6,
+  "prs": 50,
+  "missing": [],
+  "known": 28,
+  "known_of": 50,
+  "known_chance": 11.6,
+  "caught": 37,
+  "of": 47,
+  "chance": 22.6,
+  "golden": 92,
+  "golden_of": 140,
+  "flags": 914,
+  "read": 4617,
+  "flag_rate": 0.198,
+  "median_flags": 10,
+  "cost": 1.555
+ },
+ {
+  "config": "oai",
+  "label": "perch 0.4.7, OpenAI Decisions",
+  "cutoff": 0.7,
+  "prs": 50,
+  "missing": [],
+  "known": 24,
+  "known_of": 50,
+  "known_chance": 10,
+  "caught": 33,
+  "of": 47,
+  "chance": 19.8,
+  "golden": 81,
+  "golden_of": 140,
+  "flags": 760,
+  "read": 4617,
+  "flag_rate": 0.165,
+  "median_flags": 8,
+  "cost": 1.555
+ },
+ {
+  "config": "oai",
+  "label": "perch 0.4.7, OpenAI Decisions",
+  "cutoff": 0.8,
+  "prs": 50,
+  "missing": [],
+  "known": 20,
+  "known_of": 50,
+  "known_chance": 8,
+  "caught": 31,
+  "of": 47,
+  "chance": 16.6,
+  "golden": 68,
+  "golden_of": 140,
+  "flags": 596,
+  "read": 4617,
+  "flag_rate": 0.129,
+  "median_flags": 7,
+  "cost": 1.555
+ }
+]

greptile/truth/cal.com.json

@@ -0,0 +1,428 @@
+[
+ {
+  "pr": 2,
+  "index": 0,
+  "methods": [
+   "packages/app-store/_utils/getCalendar.ts::getCalendar"
+  ],
+  "file": "packages/app-store/_utils/getCalendar.ts",
+  "line": 15,
+  "why": "The new await on the dynamic appStore import has no try/catch, so a failed import rejects instead of returning null; ambiguous because the same unguarded await of appStore[...] also appears in getVideoAdapters, handlePayment, deletePayment and confirm, and getCalendar is the most central one."
+ },
+ {
+  "pr": 2,
+  "index": 1,
+  "methods": [
+   "packages/app-store/vital/lib/reschedule.ts::Reschedule",
+   "packages/app-store/wipemycalother/lib/reschedule.ts::Reschedule",
+   "packages/trpc/server/routers/viewer/bookings.tsx::requestReschedule",
+   "packages/features/bookings/lib/handleCancelBooking.ts::handler"
+  ],
+  "file": "packages/app-store/vital/lib/reschedule.ts",
+  "line": 125,
+  "why": "bookingRefsFiltered.forEach(async ...) now awaits getCalendar inside an unawaited async callback, so calendar/video deletions run fire-and-forget and the surrounding try/catch cannot see their rejections; the same pattern is in wipemycalother/lib/reschedule.ts:125, bookings.tsx:553 and handleCancelBooking.ts:460."
+ },
+ {
+  "pr": 3,
+  "index": 0,
+  "methods": [
+   "apps/web/components/auth/BackupCode.tsx::TwoFactor"
+  ],
+  "file": "apps/web/components/auth/BackupCode.tsx",
+  "line": 7,
+  "why": "The backup-code component in BackupCode.tsx is exported as a function named TwoFactor, duplicating the TwoFactor component's name."
+ },
+ {
+  "pr": 3,
+  "index": 1,
+  "methods": [
+   "apps/web/pages/api/auth/two-factor/totp/disable.ts::handler"
+  ],
+  "file": "apps/web/pages/api/auth/two-factor/totp/disable.ts",
+  "line": 50,
+  "why": "The disable endpoint logs 'cannot proceed with backup code login', copied from the login path."
+ },
+ {
+  "pr": 3,
+  "index": 2,
+  "methods": [
+   "packages/features/auth/lib/next-auth-options.ts::authorize",
+   "apps/web/pages/api/auth/two-factor/totp/disable.ts::handler"
+  ],
+  "file": "packages/features/auth/lib/next-auth-options.ts",
+  "line": 144,
+  "why": "backupCodes.indexOf(credentials.backupCode.replaceAll('-', '')) is an exact, case-sensitive match against lowercase hex codes; the disable endpoint repeats it at disable.ts:61."
+ },
+ {
+  "pr": 3,
+  "index": 3,
+  "methods": [
+   "packages/features/auth/lib/next-auth-options.ts::authorize"
+  ],
+  "file": "packages/features/auth/lib/next-auth-options.ts",
+  "line": 148,
+  "why": "The code is checked against a decrypted in-memory copy, nulled and written back with a plain update, so two concurrent logins with the same backup code can both succeed."
+ },
+ {
+  "pr": 3,
+  "index": 4,
+  "methods": [
+   "apps/web/components/settings/EnableTwoFactorModal.tsx::EnableTwoFactorModal.handleSetup",
+   "apps/web/components/settings/EnableTwoFactorModal.tsx::EnableTwoFactorModal"
+  ],
+  "file": "apps/web/components/settings/EnableTwoFactorModal.tsx",
+  "line": 99,
+  "why": "URL.createObjectURL for the backup-codes download is only revoked on the next setup call; the component has no unmount cleanup that revokes backupCodesUrl."
+ },
+ {
+  "pr": 4,
+  "index": 0,
+  "methods": [
+   "packages/app-store/googlecalendar/lib/CalendarService.ts::GoogleCalendarService.createEvent",
+   "packages/app-store/googlecalendar/lib/CalendarService.ts::GoogleCalendarService.updateEvent",
+   "packages/core/EventManager.ts::EventManager.create"
+  ],
+  "file": "packages/app-store/googlecalendar/lib/CalendarService.ts",
+  "line": 100,
+  "why": "mainHostDestinationCalendar is destructured from an empty/absent destinationCalendar array and is undefined (Google createEvent:100, updateEvent:198, as the comment says); there its uses are optional-chained, and the actual unguarded dereference is EventManager.create at EventManager.ts:119 (mainHostDestinationCalendar.integration)."
+ },
+ {
+  "pr": 4,
+  "index": 1,
+  "methods": [
+   "packages/emails/src/templates/BrokenIntegrationEmail.tsx::BrokenIntegrationEmail"
+  ],
+  "file": "packages/emails/src/templates/BrokenIntegrationEmail.tsx",
+  "line": 90,
+  "why": "mainHostDestinationCalendar?.integration uses optional chaining inside the branch of a ternary that already checked mainHostDestinationCalendar."
+ },
+ {
+  "pr": 4,
+  "index": 2,
+  "methods": [
+   "packages/app-store/googlecalendar/lib/CalendarService.ts::GoogleCalendarService.updateEvent",
+   "packages/app-store/googlecalendar/lib/CalendarService.ts::GoogleCalendarService.deleteEvent"
+  ],
+  "file": "packages/app-store/googlecalendar/lib/CalendarService.ts",
+  "line": 256,
+  "why": "In the branch where externalCalendarId is falsy, it searches destinationCalendar for externalId === externalCalendarId, which can never match; deleteEvent repeats this at line 317."
+ },
+ {
+  "pr": 4,
+  "index": 3,
+  "methods": [
+   "packages/trpc/server/routers/viewer/organizations/create.handler.ts::createHandler"
+  ],
+  "file": "packages/trpc/server/routers/viewer/organizations/create.handler.ts",
+  "line": 151,
+  "why": "The slug spread changed from !IS_TEAM_BILLING_ENABLED to IS_TEAM_BILLING_ENABLED, so with billing enabled both slug and requestedSlug are set and with billing disabled neither is."
+ },
+ {
+  "pr": 4,
+  "index": 4,
+  "methods": [
+   "packages/app-store/larkcalendar/lib/CalendarService.ts::LarkCalendarService.createEvent",
+   "packages/app-store/office365calendar/lib/CalendarService.ts::Office365CalendarService.createEvent"
+  ],
+  "file": "packages/app-store/larkcalendar/lib/CalendarService.ts",
+  "line": 125,
+  "why": "The Calendar interface (packages/types/Calendar.d.ts:221, a type declaration) now declares createEvent(event, credentialId), but the Lark (line 125) and Office365 (line 72) implementations still declare createEvent(event)."
+ },
+ {
+  "pr": 4,
+  "index": 5,
+  "methods": [
+   "packages/core/EventManager.ts::EventManager.updateAllCalendarEvents"
+  ],
+  "file": "packages/core/EventManager.ts",
+  "line": 542,
+  "why": "When reference.credentialId is set but the credential is neither in calendarCredentials nor found in the DB, the undefined credential is still passed to updateEvent."
+ },
+ {
+  "pr": 5,
+  "index": 0,
+  "methods": [
+   "packages/lib/server/service/insightsBooking.ts::InsightsBookingService.getBaseConditions",
+   "packages/lib/server/service/insightsBooking.ts::InsightsBookingService.getAuthorizationConditions"
+  ],
+  "file": "packages/lib/server/service/insightsBooking.ts",
+  "line": 76,
+  "why": "getAuthorizationConditions always returns a Prisma.Sql (NOTHING_CONDITION at worst), so authConditions is always truthy and the filter-only and NOTHING_CONDITION branches of getBaseConditions are unreachable."
+ },
+ {
+  "pr": 5,
+  "index": 1,
+  "methods": [
+   "packages/lib/server/service/insightsBooking.ts::InsightsBookingService.buildOrgAuthorizationCondition"
+  ],
+  "file": "packages/lib/server/service/insightsBooking.ts",
+  "line": 158,
+  "why": "Org member user IDs are only fetched when teamsFromOrg.length > 0, so an org with no child teams loses its members' personal bookings even though teamIds includes orgId."
+ },
+ {
+  "pr": 6,
+  "index": 0,
+  "methods": [
+   "packages/features/bookings/lib/handleCancelBooking.ts::handler",
+   "packages/features/bookings/lib/handleNewBooking.ts::handler",
+   "packages/trpc/server/routers/viewer/bookings.tsx::requestReschedule",
+   "packages/trpc/server/routers/viewer/workflows.tsx::delete",
+   "packages/trpc/server/routers/viewer/workflows.tsx::update"
+  ],
+  "file": "packages/features/bookings/lib/handleCancelBooking.ts",
+  "line": 488,
+  "why": "deleteScheduledEmailReminder/deleteScheduledSMSReminder are called without await inside forEach (handleCancelBooking:488, handleNewBooking:966, bookings.tsx:489, workflows.tsx:213/377/520/575), so cancellations may not complete."
+ },
+ {
+  "pr": 6,
+  "index": 1,
+  "methods": [
+   "packages/features/ee/workflows/lib/reminders/emailReminderManager.ts::deleteScheduledEmailReminder"
+  ],
+  "file": "packages/features/ee/workflows/lib/reminders/emailReminderManager.ts",
+  "line": 213,
+  "why": "The immediateDelete branch cancels the SendGrid batch and returns without deleting the WorkflowReminder row or calling the scheduled_sends DELETE."
+ },
+ {
+  "pr": 6,
+  "index": 2,
+  "methods": [],
+  "file": "packages/prisma/schema.prisma",
+  "line": 644,
+  "why": "WorkflowReminder.cancelled is added as Boolean? with no default (migration adds a nullable BOOLEAN), so existing rows are null instead of false; schema/migration, not a method."
+ },
+ {
+  "pr": 7,
+  "index": 0,
+  "methods": [
+   "packages/trpc/server/routers/viewer/slots.ts::checkIfIsAvailable"
+  ],
+  "file": "packages/trpc/server/routers/viewer/slots.ts",
+  "line": 142,
+  "why": "The working-hours check computes end from slotStartTime instead of slotEndTime, so slots that run past the end of working hours pass."
+ },
+ {
+  "pr": 7,
+  "index": 1,
+  "methods": [
+   "packages/trpc/server/routers/viewer/slots.ts::checkIfIsAvailable"
+  ],
+  "file": "packages/trpc/server/routers/viewer/slots.ts",
+  "line": 114,
+  "why": "Two dayjs objects are compared with ===, which compares references and is always false; it should use isSame."
+ },
+ {
+  "pr": 8,
+  "index": 0,
+  "methods": [
+   "packages/app-store/_utils/oauth/parseRefreshTokenResponse.ts::parseRefreshTokenResponse"
+  ],
+  "file": "packages/app-store/_utils/oauth/parseRefreshTokenResponse.ts",
+  "line": 26,
+  "why": "A missing refresh_token is replaced with the literal string 'refresh_token', which is then persisted and breaks later refreshes."
+ },
+ {
+  "pr": 8,
+  "index": 1,
+  "methods": [
+   "packages/app-store/_utils/oauth/parseRefreshTokenResponse.ts::<top-level>"
+  ],
+  "file": "packages/app-store/_utils/oauth/parseRefreshTokenResponse.ts",
+  "line": 8,
+  "why": "minimumTokenResponseSchema uses computed keys [z.string().toString()] and [z.string().optional().toString()], which become literal '[object Object]'-style keys, not wildcards; module-level const (used by parseRefreshTokenResponse at line 16), not inside a listed method. (mapped to the file's <top-level> unit, which perch reads)"
+ },
+ {
+  "pr": 8,
+  "index": 2,
+  "methods": [
+   "packages/app-store/googlecalendar/lib/CalendarService.ts::GoogleCalendarService.<anonymous>.refreshAccessToken",
+   "packages/app-store/_utils/oauth/parseRefreshTokenResponse.ts::parseRefreshTokenResponse"
+  ],
+  "file": "packages/app-store/googlecalendar/lib/CalendarService.ts",
+  "line": 97,
+  "why": "key = parseRefreshTokenResponse(...) is the safeParse wrapper ({success, data}) and is persisted as the credential key instead of .data."
+ },
+ {
+  "pr": 8,
+  "index": 3,
+  "methods": [
+   "packages/app-store/_utils/oauth/refreshOAuthTokens.ts::refreshOAuthTokens",
+   "packages/app-store/googlecalendar/lib/CalendarService.ts::GoogleCalendarService.<anonymous>.refreshAccessToken",
+   "packages/app-store/hubspot/lib/CalendarService.ts::HubspotCalendarService.<anonymous>.refreshAccessToken"
+  ],
+  "file": "packages/app-store/_utils/oauth/refreshOAuthTokens.ts",
+  "line": 15,
+  "why": "In the sync-endpoint branch refreshOAuthTokens returns the raw fetch Response, while callers such as Google (res.data) and Hubspot (HubspotToken) expect the integration's token object."
+ },
+ {
+  "pr": 8,
+  "index": 4,
+  "methods": [
+   "packages/app-store/googlecalendar/lib/CalendarService.ts::GoogleCalendarService.<anonymous>.refreshAccessToken",
+   "packages/app-store/_utils/oauth/refreshOAuthTokens.ts::refreshOAuthTokens"
+  ],
+  "file": "packages/app-store/googlecalendar/lib/CalendarService.ts",
+  "line": 94,
+  "why": "On the sync path res is a fetch Response, so res?.data is undefined and token.access_token on line 95 throws."
+ },
+ {
+  "pr": 8,
+  "index": 5,
+  "methods": [
+   "packages/app-store/zoho-bigin/lib/CalendarService.ts::BiginCalendarService.refreshAccessToken"
+  ],
+  "file": "packages/app-store/zoho-bigin/lib/CalendarService.ts",
+  "line": 93,
+  "why": "refreshOAuthTokens is passed credentialId where it expects the userId, so the sync endpoint receives the wrong calcomUserId."
+ },
+ {
+  "pr": 8,
+  "index": 6,
+  "methods": [
+   "packages/app-store/salesforce/lib/CalendarService.ts::<top-level>"
+  ],
+  "file": "packages/app-store/salesforce/lib/CalendarService.ts",
+  "line": 86,
+  "why": "The Salesforce token refresh checks response.statusText !== 'OK' instead of response.ok; this is in the getClient class-property arrow function, which methods_read does not list. (mapped to the file's <top-level> unit, which perch reads)"
+ },
+ {
+  "pr": 8,
+  "index": 7,
+  "methods": [
+   "packages/app-store/salesforce/lib/CalendarService.ts::<top-level>"
+  ],
+  "file": "packages/app-store/salesforce/lib/CalendarService.ts",
+  "line": 106,
+  "why": "After refreshing and saving new tokens, jsforce.Connection is still built from the stale credentialKey.access_token/instance_url (lines 105-106); this is in getClient, which methods_read does not list. (mapped to the file's <top-level> unit, which perch reads)"
+ },
+ {
+  "pr": 8,
+  "index": 8,
+  "methods": [
+   "apps/web/pages/api/webhook/app-credential.ts::handler"
+  ],
+  "file": "apps/web/pages/api/webhook/app-credential.ts",
+  "line": 31,
+  "why": "appCredentialWebhookRequestBodySchema.parse throws on an invalid body, producing a 500 instead of a 400."
+ },
+ {
+  "pr": 9,
+  "index": 0,
+  "methods": [
+   "packages/features/ee/workflows/api/scheduleSMSReminders.ts::handler"
+  ],
+  "file": "packages/features/ee/workflows/api/scheduleSMSReminders.ts",
+  "line": 184,
+  "why": "retryCount: reminder.retryCount + 1 writes a value read earlier instead of an atomic increment; the catch block repeats it at line 195."
+ },
+ {
+  "pr": 9,
+  "index": 1,
+  "methods": [
+   "packages/features/ee/workflows/api/scheduleSMSReminders.ts::handler"
+  ],
+  "file": "packages/features/ee/workflows/api/scheduleSMSReminders.ts",
+  "line": 38,
+  "why": "The OR branch { retryCount: { gt: 1 } } has no method: SMS filter, so deleteMany removes email/WhatsApp reminders with retryCount > 1."
+ },
+ {
+  "pr": 10,
+  "index": 0,
+  "methods": [
+   "packages/trpc/server/routers/viewer/bookings/addGuests.handler.ts::addGuestsHandler"
+  ],
+  "file": "packages/trpc/server/routers/viewer/bookings/addGuests.handler.ts",
+  "line": 77,
+  "why": "The blacklist is lowercased but guest emails are not, so blacklistedGuestEmails.includes(guest) is bypassed by uppercase input."
+ },
+ {
+  "pr": 10,
+  "index": 1,
+  "methods": [
+   "packages/trpc/server/routers/viewer/bookings/addGuests.handler.ts::addGuestsHandler"
+  ],
+  "file": "packages/trpc/server/routers/viewer/bookings/addGuests.handler.ts",
+  "line": 47,
+  "why": "isTeamAdminOrOwner requires isTeamAdmin && isTeamOwner; it should be ||."
+ },
+ {
+  "pr": 10,
+  "index": 2,
+  "methods": [
+   "packages/emails/email-manager.ts::sendAddGuestsEmails",
+   "packages/trpc/server/routers/viewer/bookings/addGuests.handler.ts::addGuestsHandler"
+  ],
+  "file": "packages/emails/email-manager.ts",
+  "line": 541,
+  "why": "sendAddGuestsEmails decides which attendees are new with newGuests.includes(attendee.email), but the handler passes the raw input guests (addGuests.handler.ts:168), not the filtered uniqueGuests."
+ },
+ {
+  "pr": 10,
+  "index": 3,
+  "methods": [
+   "packages/trpc/server/routers/viewer/bookings/addGuests.handler.ts::addGuestsHandler"
+  ],
+  "file": "packages/trpc/server/routers/viewer/bookings/addGuests.handler.ts",
+  "line": 74,
+  "why": "uniqueGuests removes existing attendees and blacklisted emails but not duplicates within the input, so createMany can insert duplicate attendees."
+ },
+ {
+  "pr": 10,
+  "index": 4,
+  "methods": [
+   "apps/web/components/dialog/AddGuestsDialog.tsx::AddGuestsDialog"
+  ],
+  "file": "apps/web/components/dialog/AddGuestsDialog.tsx",
+  "line": 32,
+  "why": "multiEmailValue starts as [\"\"] (and is reset to it), so an empty string is submitted and fails email validation."
+ },
+ {
+  "pr": 10,
+  "index": 5,
+  "methods": [
+   "packages/emails/email-manager.ts::sendAddGuestsEmails"
+  ],
+  "file": "packages/emails/email-manager.ts",
+  "line": 529,
+  "why": "sendAddGuestsEmails sends organizer, team and attendee emails without checking the event type's disableStandardEmails metadata."
+ },
+ {
+  "pr": 11,
+  "index": 0,
+  "methods": [
+   "packages/app-store/googlecalendar/lib/CalendarService.ts::GoogleCalendarService.fetchAvailabilityAndSetCache",
+   "packages/lib/server/repository/selectedCalendar.ts::SelectedCalendarRepository.updateManyByCredentialId"
+  ],
+  "file": "packages/app-store/googlecalendar/lib/CalendarService.ts",
+  "line": 1024,
+  "why": "updateManyByCredentialId(this.credential.id, {}) passes empty data, so updatedAt is not reliably bumped and the cache status timestamp is stale."
+ },
+ {
+  "pr": 11,
+  "index": 1,
+  "methods": [],
+  "file": "scripts/test-gcal-webhooks.sh",
+  "line": 68,
+  "why": "sed -i '' is BSD/macOS syntax and fails with GNU sed on Linux; shell script, not a method."
+ },
+ {
+  "pr": 11,
+  "index": 2,
+  "methods": [
+   "packages/trpc/server/routers/viewer/calendars/deleteCache.handler.ts::deleteCacheHandler"
+  ],
+  "file": "packages/trpc/server/routers/viewer/calendars/deleteCache.handler.ts",
+  "line": 25,
+  "why": "A plain Error is thrown for a missing or unauthorized credential, which tRPC returns as a 500 instead of a TRPCError FORBIDDEN/NOT_FOUND."
+ },
+ {
+  "pr": 11,
+  "index": 3,
+  "methods": [
+   "packages/features/apps/components/CredentialActionsDropdown.tsx::CredentialActionsDropdown"
+  ],
+  "file": "packages/features/apps/components/CredentialActionsDropdown.tsx",
+  "line": 89,
+  "why": "The cache timestamp is formatted with a hardcoded 'en-US' Intl.DateTimeFormat instead of the user's locale."
+ }
+]

greptile/truth/discourse.json

@@ -0,0 +1,390 @@
+[
+ {
+  "pr": 1,
+  "index": 0,
+  "methods": [
+   "app/models/optimized_image.rb::OptimizedImage.downsize#2",
+   "app/models/optimized_image.rb::OptimizedImage.downsize"
+  ],
+  "file": "app/models/optimized_image.rb",
+  "line": 149,
+  "why": "A second self.downsize(from, to, dimensions, opts) is defined right after the four-argument one at line 145, so Ruby keeps only the three-argument version and any caller passing max_width and max_height separately gets the wrong arguments."
+ },
+ {
+  "pr": 1,
+  "index": 1,
+  "methods": [
+   "app/assets/javascripts/discourse/lib/utilities.js::Discourse.Utilities.validateUploadedFile",
+   "app/assets/javascripts/discourse/lib/utilities.js::Discourse.Utilities.displayErrorForUpload"
+  ],
+  "file": "app/assets/javascripts/discourse/lib/utilities.js",
+  "line": 182,
+  "why": "validateUploadedFile hardcodes maxSizeKB = 10 * 1024 instead of reading Discourse.SiteSettings['max_' + type + '_size_kb']; the 413 branch of displayErrorForUpload (line 246) does the same."
+ },
+ {
+  "pr": 1,
+  "index": 2,
+  "methods": [
+   "app/controllers/uploads_controller.rb::UploadsController.create_upload",
+   "app/models/optimized_image.rb::OptimizedImage.resize_instructions_animated"
+  ],
+  "file": "app/controllers/uploads_controller.rb",
+  "line": 67,
+  "why": "create_upload passes \"80%\" as the dimensions with allow_animation set; for GIFs downsize_instructions_animated delegates to resize_instructions_animated, whose gifsicle --resize-fit (line 120) expects WxH, so the downsize fails silently."
+ },
+ {
+  "pr": 1,
+  "index": 3,
+  "methods": [
+   "app/controllers/uploads_controller.rb::UploadsController.create_upload"
+  ],
+  "file": "app/controllers/uploads_controller.rb",
+  "line": 66,
+  "why": "The while loop tests tempfile.size after convert has rewritten the file at tempfile.path, so the size read through the open Tempfile can be stale and the loop does not reliably see the shrink."
+ },
+ {
+  "pr": 2,
+  "index": 0,
+  "methods": [
+   "app/controllers/topics_controller.rb::TopicsController.unsubscribe"
+  ],
+  "file": "app/controllers/topics_controller.rb",
+  "line": 107,
+  "why": "TopicUser.find_by on line 105 returns nil when the user has no TopicUser row for the topic, and line 107 calls tu.notification_level on it."
+ },
+ {
+  "pr": 2,
+  "index": 1,
+  "methods": [],
+  "file": "app/assets/javascripts/discourse/controllers/topic-unsubscribe.js.es6",
+  "line": 5,
+  "why": "The computed property is spelled stopNotificiationsText (also used that way in templates/topic/unsubscribe.hbs line 3); it is in an ES6 controller that methods_read does not list, and the misspelling is consistent so it is cosmetic."
+ },
+ {
+  "pr": 3,
+  "index": 0,
+  "methods": [
+   "app/models/blocked_email.rb::BlockedEmail.should_block?"
+  ],
+  "file": "app/models/blocked_email.rb",
+  "line": 14,
+  "why": "should_block? increments match_count, sets last_match_at and saves the record (lines 14-16) every time it is asked, so a validation check writes to the database with a non-atomic read-modify-write."
+ },
+ {
+  "pr": 3,
+  "index": 1,
+  "methods": [
+   "lib/validators/email_validator.rb::EmailValidator.email_in_restriction_setting?"
+  ],
+  "file": "lib/validators/email_validator.rb",
+  "line": 20,
+  "why": "Regexp.new(\"@(#{domains})\", true) has no end anchor or boundary, so the whitelist/blacklist match accepts addresses whose domain only contains the listed domain rather than equalling it."
+ },
+ {
+  "pr": 3,
+  "index": 2,
+  "methods": [
+   "app/models/blocked_email.rb::BlockedEmail.should_block?"
+  ],
+  "file": "app/models/blocked_email.rb",
+  "line": 12,
+  "why": "BlockedEmail.where(email: email) compares case-sensitively, so a differently-cased address bypasses a block."
+ },
+ {
+  "pr": 4,
+  "index": 0,
+  "methods": [
+   "app/jobs/scheduled/poll_feed.rb::Jobs.PollFeed.poll_feed"
+  ],
+  "file": "app/jobs/scheduled/poll_feed.rb",
+  "line": 29,
+  "why": "poll_feed calls Kernel#open on SiteSetting.feed_polling_url with no scheme or host validation, so the server fetches (or open-uri pipes) whatever URL the setting holds."
+ },
+ {
+  "pr": 4,
+  "index": 1,
+  "methods": [
+   "app/assets/javascripts/embed.js::<anonymous>.postMessageReceived"
+  ],
+  "file": "app/assets/javascripts/embed.js",
+  "line": 17,
+  "why": "discourseUrl.indexOf(e.origin) is a substring test, not an exact origin comparison, so other origins that appear inside discourseUrl pass the check."
+ },
+ {
+  "pr": 4,
+  "index": 2,
+  "methods": [],
+  "file": "app/views/layouts/embed.html.erb",
+  "line": 11,
+  "why": "In the layout template (no method), postMessage uses the full request.referer URL as targetOrigin instead of the referer's scheme+host+port."
+ },
+ {
+  "pr": 4,
+  "index": 3,
+  "methods": [
+   "app/controllers/embed_controller.rb::EmbedController.ensure_embeddable"
+  ],
+  "file": "app/controllers/embed_controller.rb",
+  "line": 28,
+  "why": "ensure_embeddable sets X-Frame-Options to ALLOWALL after only a Referer host check (line 26), which also maps a missing referer to ''."
+ },
+ {
+  "pr": 4,
+  "index": 4,
+  "methods": [
+   "app/models/topic_embed.rb::TopicEmbed.import"
+  ],
+  "file": "app/models/topic_embed.rb",
+  "line": 13,
+  "why": "import appends to contents with << (NoMethodError when contents is nil, and it mutates the caller's string) and interpolates url unescaped into the <a href> HTML."
+ },
+ {
+  "pr": 4,
+  "index": 5,
+  "methods": [],
+  "file": "app/views/embed/best.html.erb",
+  "line": 6,
+  "why": "The ERB template closes the if/else with '<%- end if %>' instead of '<%- end %>'; it is in a view template, not a method."
+ },
+ {
+  "pr": 4,
+  "index": 6,
+  "methods": [
+   "app/jobs/scheduled/poll_feed.rb::Jobs.PollFeed.poll_feed"
+  ],
+  "file": "app/jobs/scheduled/poll_feed.rb",
+  "line": 35,
+  "why": "i.content.scrub raises NoMethodError for feed items that have no content element; there is no fallback to i.description."
+ },
+ {
+  "pr": 4,
+  "index": 7,
+  "methods": [],
+  "file": "app/views/layouts/embed.html.erb",
+  "line": 11,
+  "why": "In the layout template (no method), request.referer is interpolated into a JavaScript string literal with only HTML escaping, not JavaScript escaping."
+ },
+ {
+  "pr": 5,
+  "index": 0,
+  "methods": [],
+  "file": "app/assets/stylesheets/common/base/topic.scss",
+  "line": 34,
+  "why": "SCSS, no methods: .extra-info-wrapper becomes a flex item (order(2)) but its .badge-wrapper keeps float: left; the PR also drops float: right from .d-header .panel (header.scss lines 35-39), relying on margin-left: auto, which only works when .panel is a flex item."
+ },
+ {
+  "pr": 5,
+  "index": 1,
+  "methods": [],
+  "file": "app/assets/stylesheets/common/foundation/mixins.scss",
+  "line": 121,
+  "why": "SCSS, no methods: the align-items mixin emits -ms-align-items, which is not a real property; -ms-flex-align (line 120) is the IE one."
+ },
+ {
+  "pr": 5,
+  "index": 2,
+  "methods": [],
+  "file": "app/assets/stylesheets/common/foundation/mixins.scss",
+  "line": 126,
+  "why": "SCSS, no methods: the order mixin passes $val straight to -webkit-box-ordinal-group and -moz-box-ordinal-group (lines 126-127), which start at 1, so they do not get the same ordering as order: $val."
+ },
+ {
+  "pr": 6,
+  "index": 0,
+  "methods": [
+   "app/serializers/user_serializer.rb::UserSerializer.include_website_name",
+   "app/serializers/user_serializer.rb::UserSerializer.website_name"
+  ],
+  "file": "app/serializers/user_serializer.rb",
+  "line": 153,
+  "why": "The conditional is named include_website_name without the trailing ?, so ActiveModel::Serializers never treats it as the include check; website_name also builds \".\" << website_host (line 149), mutating a string literal."
+ },
+ {
+  "pr": 7,
+  "index": 0,
+  "methods": [],
+  "file": "app/assets/stylesheets/desktop/topic-post.scss",
+  "line": 291,
+  "why": "SCSS, no methods: .topic-meta-data h5 a went from scale-color($primary, 30%) to dark-light-choose(scale-color($primary, 70%), ...), so the light-theme value is inverted rather than kept."
+ },
+ {
+  "pr": 7,
+  "index": 1,
+  "methods": [],
+  "file": "app/assets/stylesheets/desktop/user.scss",
+  "line": 522,
+  "why": "SCSS, no methods: .group-member-info .name changes the light-theme $primary lightness from 30% to 50% instead of keeping 30% (same change in mobile/user.scss line 497)."
+ },
+ {
+  "pr": 7,
+  "index": 2,
+  "methods": [],
+  "file": "app/assets/stylesheets/desktop/topic-post.scss",
+  "line": 291,
+  "why": "SCSS, no methods. Ambiguous: no topic-post file has a light-theme 70% -> 30% change; the only topic-post line where a 30/70 pair was swapped is desktop/topic-post.scss 291 (30% -> 70%). The only literal 70% -> 30% change is mobile/modal.scss line 102 (.custom-message-length)."
+ },
+ {
+  "pr": 7,
+  "index": 3,
+  "methods": [],
+  "file": "app/assets/stylesheets/desktop/topic-post.scss",
+  "line": 291,
+  "why": "SCSS, no methods: same line as golden 0; the light-theme lightness for .topic-meta-data h5 a moved from 30% to 70%, which lowers contrast."
+ },
+ {
+  "pr": 7,
+  "index": 4,
+  "methods": [],
+  "file": "app/assets/stylesheets/mobile/user.scss",
+  "line": 497,
+  "why": "SCSS, no methods: .group-member-info .name became dark-light-choose(50%, 50%) instead of keeping $primary 30% with the $secondary 70% complement."
+ },
+ {
+  "pr": 8,
+  "index": 0,
+  "methods": [
+   "app/assets/javascripts/admin/routes/admin_group_route.js::setupController"
+  ],
+  "file": "app/assets/javascripts/admin/routes/admin_group_route.js",
+  "line": 13,
+  "why": "afterModel was removed and setupController now calls model.findMembers() without returning or waiting for its promise, so the page renders before members load (same in routes/group-members.js.es6 line 11, which methods_read does not list)."
+ },
+ {
+  "pr": 8,
+  "index": 1,
+  "methods": [],
+  "file": "app/assets/javascripts/admin/controllers/admin-group.js.es6",
+  "line": 33,
+  "why": "The next action caps offset at user_count with Math.min, so offset can equal the total and the page fetched is empty; it is in the ES6 admin-group controller, which methods_read does not list."
+ },
+ {
+  "pr": 8,
+  "index": 2,
+  "methods": [],
+  "file": "spec/controllers/admin/groups_controller_spec.rb",
+  "line": 115,
+  "why": "The spec calls xhr :put, :remove_member although the route is delete \"members\"; spec code, not in methods_read."
+ },
+ {
+  "pr": 8,
+  "index": 3,
+  "methods": [],
+  "file": "app/assets/javascripts/admin/controllers/admin-group.js.es6",
+  "line": 13,
+  "why": "totalPages is Math.floor(user_count / limit) + 1, which adds a page when user_count is an exact multiple of limit; it is in the ES6 admin-group controller, which methods_read does not list."
+ },
+ {
+  "pr": 8,
+  "index": 4,
+  "methods": [
+   "app/assets/javascripts/discourse/models/group.js::<top-level>"
+  ],
+  "file": "app/assets/javascripts/discourse/models/group.js",
+  "line": 17,
+  "why": "userCountDisplay (in the Discourse.Group extend body that the <top-level> entry covers) returns c only when c > 0 and falls off the end otherwise. Ambiguous: findMembers line 21 now does a bare 'return ;' where it used to return a resolved promise."
+ },
+ {
+  "pr": 8,
+  "index": 5,
+  "methods": [
+   "app/controllers/admin/groups_controller.rb::Admin::GroupsController.add_members",
+   "app/controllers/admin/groups_controller.rb::Admin::GroupsController.remove_member"
+  ],
+  "file": "app/controllers/admin/groups_controller.rb",
+  "line": 66,
+  "why": "add_members (line 66) and remove_member (line 85) read params.require(:group_id); the comment says the route supplies :id. Rails routes declared inside a resources block normally pass :group_id, so the claim is doubtful, but these are the lines it refers to."
+ },
+ {
+  "pr": 9,
+  "index": 0,
+  "methods": [
+   "lib/freedom_patches/translate_accelerator.rb::I18n.ensure_loaded!"
+  ],
+  "file": "lib/freedom_patches/translate_accelerator.rb",
+  "line": 63,
+  "why": "ensure_loaded! lazily initializes @loaded_locales with ||= and reads it outside LOAD_MUTEX, with no synchronization between threads."
+ },
+ {
+  "pr": 9,
+  "index": 1,
+  "methods": [
+   "lib/freedom_patches/translate_accelerator.rb::I18n.ensure_loaded!",
+   "config/initializers/i18n.rb::FallbackLocaleList.[]"
+  ],
+  "file": "lib/freedom_patches/translate_accelerator.rb",
+  "line": 64,
+  "why": "ensure_loaded! checks @loaded_locales.include?(locale) without normalizing to a symbol, so 'en' and :en load twice; FallbackLocaleList#[] (i18n.rb line 17) mixes the raw locale with symbols, so uniq does not dedupe them."
+ },
+ {
+  "pr": 10,
+  "index": 0,
+  "methods": [
+   "app/models/embeddable_host.rb::<top-level>"
+  ],
+  "file": "app/models/embeddable_host.rb",
+  "line": 6,
+  "why": "The class-level before_validation block calls self.host.sub! without a nil guard, so NoMethodError is raised when host is nil."
+ },
+ {
+  "pr": 10,
+  "index": 1,
+  "methods": [
+   "app/controllers/admin/embeddable_hosts_controller.rb::Admin::EmbeddableHostsController.update",
+   "app/controllers/admin/embeddable_hosts_controller.rb::Admin::EmbeddableHostsController.destroy"
+  ],
+  "file": "app/controllers/admin/embeddable_hosts_controller.rb",
+  "line": 10,
+  "why": "update (line 10) and destroy (line 15) use EmbeddableHost.where(id:).first and then call save_host or host.destroy on a possibly nil record."
+ },
+ {
+  "pr": 10,
+  "index": 2,
+  "methods": [
+   "app/models/embeddable_host.rb::EmbeddableHost.record_for_host"
+  ],
+  "file": "app/models/embeddable_host.rb",
+  "line": 17,
+  "why": "record_for_host compares lower(host) = ? with uri.host as given, without downcasing, so mixed-case referer hosts do not match."
+ },
+ {
+  "pr": 10,
+  "index": 3,
+  "methods": [
+   "db/migrate/20150818190757_create_embeddable_hosts.rb::CreateEmbeddableHosts.change"
+  ],
+  "file": "db/migrate/20150818190757_create_embeddable_hosts.rb",
+  "line": 25,
+  "why": "The migration inserts each old embeddable_hosts value with raw SQL, skipping the model's scheme/path stripping, so migrated hosts with http:// or a path never match record_for_host."
+ },
+ {
+  "pr": 10,
+  "index": 4,
+  "methods": [
+   "spec/fabricators/category_fabricator.rb::<top-level>",
+   "spec/fabricators/embeddable_host_fabricator.rb::<top-level>"
+  ],
+  "file": "spec/fabricators/category_fabricator.rb",
+  "line": 1,
+  "why": "category_fabricator.rb now holds the :embeddable_host fabricator and embeddable_host_fabricator.rb holds the category fabricators; the contents are swapped between the two files."
+ },
+ {
+  "pr": 10,
+  "index": 5,
+  "methods": [
+   "db/migrate/20150818190757_create_embeddable_hosts.rb::CreateEmbeddableHosts.change"
+  ],
+  "file": "db/migrate/20150818190757_create_embeddable_hosts.rb",
+  "line": 25,
+  "why": "The '#{h}' interpolation is in the migration's INSERT, not in the controller the comment names; host values are put into SQL unquoted."
+ },
+ {
+  "pr": 10,
+  "index": 6,
+  "methods": [
+   "db/migrate/20150818190757_create_embeddable_hosts.rb::CreateEmbeddableHosts.change"
+  ],
+  "file": "db/migrate/20150818190757_create_embeddable_hosts.rb",
+  "line": 11,
+  "why": "execute(...)[0]['id'] (statement starting line 9) indexes the first row with no check, which raises when no embed_category setting row exists; line 15 does the same for uncategorized_category_id."
+ }
+]

greptile/truth/grafana.json

@@ -0,0 +1,261 @@
+[
+ {
+  "pr": 1,
+  "index": 0,
+  "methods": [
+   "pkg/services/anonymous/anonimpl/anonstore/database.go::AnonDBStore.CreateOrUpdateDevice"
+  ],
+  "file": "pkg/services/anonymous/anonimpl/anonstore/database.go",
+  "line": 110,
+  "why": "CountDevices is read and the insert happens afterwards with no transaction or lock, so concurrent requests can all pass the count >= deviceLimit check and exceed the limit."
+ },
+ {
+  "pr": 1,
+  "index": 1,
+  "methods": [
+   "pkg/services/anonymous/anonimpl/client.go::Anonymous.Authenticate",
+   "pkg/services/anonymous/anonimpl/impl.go::AnonDeviceService.TagDevice"
+  ],
+  "file": "pkg/services/anonymous/anonimpl/client.go",
+  "line": 44,
+  "why": "Authenticate now calls TagDevice synchronously and returns ErrDeviceLimitReached as an auth failure (lines 44-47); TagDevice was changed to propagate the tagging error (impl.go line 147)."
+ },
+ {
+  "pr": 1,
+  "index": 2,
+  "methods": [
+   "pkg/services/anonymous/anonimpl/anonstore/database.go::AnonDBStore.updateDevice"
+  ],
+  "file": "pkg/services/anonymous/anonimpl/anonstore/database.go",
+  "line": 95,
+  "why": "rowsAffected == 0 is mapped to ErrDeviceLimitReached, although it only means no existing device matched the id/time window."
+ },
+ {
+  "pr": 1,
+  "index": 3,
+  "methods": [
+   "pkg/services/anonymous/anonimpl/anonstore/database.go::AnonDBStore.updateDevice"
+  ],
+  "file": "pkg/services/anonymous/anonimpl/anonstore/database.go",
+  "line": 81,
+  "why": "The BETWEEN bounds are computed from device.UpdatedAt rather than time.Now(), unlike the CountDevices window in CreateOrUpdateDevice."
+ },
+ {
+  "pr": 1,
+  "index": 4,
+  "methods": [
+   "pkg/services/anonymous/anonimpl/impl.go::AnonDeviceService.tagDeviceUI"
+  ],
+  "file": "pkg/services/anonymous/anonimpl/impl.go",
+  "line": 87,
+  "why": "localCache.SetDefault runs before CreateOrUpdateDevice, so a device rejected with ErrDeviceLimitReached stays cached and skips the limit check for 29 minutes."
+ },
+ {
+  "pr": 2,
+  "index": 0,
+  "methods": [
+   "pkg/services/authz/rbac/service.go::Service.Check"
+  ],
+  "file": "pkg/services/authz/rbac/service.go",
+  "line": 131,
+  "why": "Cached permissions are trusted only when they allow (lines 123-136); a cached deny falls through to a fresh DB lookup, so a stale cached grant still allows access after revocation."
+ },
+ {
+  "pr": 2,
+  "index": 1,
+  "methods": [],
+  "file": "pkg/services/authz/rbac/service_test.go",
+  "line": 981,
+  "why": "Test file (not in methods_read): the comment says the cached permissions allow access but the map stores false for dashboards:uid:dash1."
+ },
+ {
+  "pr": 2,
+  "index": 2,
+  "methods": [
+   "pkg/services/authz/rbac/cache.go::userPermDenialCacheKey"
+  ],
+  "file": "pkg/services/authz/rbac/cache.go",
+  "line": 31,
+  "why": "The denial key joins namespace, user, action, name and parent with underscores, so values containing underscores or empty parts can collide."
+ },
+ {
+  "pr": 3,
+  "index": 0,
+  "methods": [
+   "pkg/services/pluginsintegration/clientmiddleware/contextual_logger_middleware.go::ContextualLoggerMiddleware.QueryData",
+   "pkg/services/pluginsintegration/clientmiddleware/contextual_logger_middleware.go::ContextualLoggerMiddleware.CallResource",
+   "pkg/services/pluginsintegration/clientmiddleware/contextual_logger_middleware.go::ContextualLoggerMiddleware.CheckHealth",
+   "pkg/services/pluginsintegration/clientmiddleware/contextual_logger_middleware.go::ContextualLoggerMiddleware.CollectMetrics"
+  ],
+  "file": "pkg/services/pluginsintegration/clientmiddleware/contextual_logger_middleware.go",
+  "line": 40,
+  "why": "Each method dereferences req.PluginContext (lines 40, 45, 50, 55) with no req == nil guard, unlike LoggerMiddleware, so a nil request panics."
+ },
+ {
+  "pr": 3,
+  "index": 1,
+  "methods": [
+   "pkg/services/pluginsintegration/clientmiddleware/logger_middleware.go::LoggerMiddleware.logRequest",
+   "pkg/services/pluginsintegration/clientmiddleware/contextual_logger_middleware.go::instrumentContext"
+  ],
+  "file": "pkg/services/pluginsintegration/clientmiddleware/logger_middleware.go",
+  "line": 49,
+  "why": "logRequest no longer adds traceID to logParams, and the new instrumentContext does not add it either, so plugin request logs lose the trace ID."
+ },
+ {
+  "pr": 3,
+  "index": 2,
+  "methods": [
+   "pkg/plugins/log/fake.go::TestLogger.FromContext"
+  ],
+  "file": "pkg/plugins/log/fake.go",
+  "line": 47,
+  "why": "FromContext returns NewTestLogger() instead of the receiver, so logs written through it are lost to test assertions."
+ },
+ {
+  "pr": 4,
+  "index": 0,
+  "methods": [
+   "public/app/plugins/datasource/loki/shardQuerySplitting.ts::runShardSplitQuery",
+   "public/app/plugins/datasource/loki/querySplitting.ts::runSplitQuery"
+  ],
+  "file": "public/app/plugins/datasource/loki/shardQuerySplitting.ts",
+  "line": 52,
+  "why": "applyTemplateVariables is called with request.filters, but the shardQuerySplitting tests mock it as a one-argument function, so the filters path is untested; the same call is in runSplitQuery line 299 (file choice is ambiguous)."
+ },
+ {
+  "pr": 5,
+  "index": 0,
+  "methods": [
+   "public/app/features/alerting/unified/rule-list/FilterView.tsx::FilterViewResults"
+  ],
+  "file": "public/app/features/alerting/unified/rule-list/FilterView.tsx",
+  "line": 157,
+  "why": "The grafana case renders GrafanaRuleListItem inside rules.map without key={key} (the other cases keep it)."
+ },
+ {
+  "pr": 5,
+  "index": 1,
+  "methods": [
+   "public/app/features/alerting/unified/rule-list/components/RuleActionsButtons.V2.tsx::RuleActionsButtons",
+   "public/app/features/alerting/unified/rule-list/GrafanaRuleListItem.tsx::GrafanaRuleListItem"
+  ],
+  "file": "public/app/features/alerting/unified/rule-list/components/RuleActionsButtons.V2.tsx",
+  "line": 104,
+  "why": "SilenceGrafanaRuleDrawer renders only when rulerRuleType.grafana.alertingRule(rule), but GrafanaRuleListItem (line 47) now passes only promRule, so 'Silence notifications' sets state and shows nothing."
+ },
+ {
+  "pr": 6,
+  "index": 0,
+  "methods": [
+   "pkg/apiserver/rest/dualwriter_mode3.go::DualWriterMode3.Delete"
+  ],
+  "file": "pkg/apiserver/rest/dualwriter_mode3.go",
+  "line": 97,
+  "why": "klog.NewContext(ctx, d.Log) uses the base logger instead of the enriched log with name/kind/method."
+ },
+ {
+  "pr": 6,
+  "index": 1,
+  "methods": [
+   "pkg/apiserver/rest/dualwriter_mode3.go::DualWriterMode3.DeleteCollection"
+  ],
+  "file": "pkg/apiserver/rest/dualwriter_mode3.go",
+  "line": 166,
+  "why": "In DeleteCollection the legacy DeleteCollection duration is recorded with recordStorageDuration instead of recordLegacyDuration; the comment's wording is garbled, and Create (line 45) and Update (line 129) have the inverse mix-up."
+ },
+ {
+  "pr": 6,
+  "index": 2,
+  "methods": [
+   "pkg/apiserver/rest/dualwriter_mode3.go::DualWriterMode3.Delete"
+  ],
+  "file": "pkg/apiserver/rest/dualwriter_mode3.go",
+  "line": 106,
+  "why": "recordStorageDuration is passed name instead of options.Kind as the kind label."
+ },
+ {
+  "pr": 7,
+  "index": 0,
+  "methods": [
+   "pkg/services/annotations/annotationsimpl/xorm_store.go::xormRepositoryImpl.CleanAnnotations",
+   "pkg/services/annotations/annotationsimpl/xorm_store.go::xormRepositoryImpl.CleanOrphanedAnnotationTags"
+  ],
+  "file": "pkg/services/annotations/annotationsimpl/xorm_store.go",
+  "line": 534,
+  "why": "Debug output (ids, counts, cond) is logged with r.log.Error on every batch (lines 534, 537, 554, 557 and 576, 579)."
+ },
+ {
+  "pr": 7,
+  "index": 1,
+  "methods": [
+   "pkg/services/cleanup/cleanup.go::CleanUpService.Run"
+  ],
+  "file": "pkg/services/cleanup/cleanup.go",
+  "line": 77,
+  "why": "The cleanup ticker changed from 10 minutes to 1 minute, running every cleanup job ten times as often."
+ },
+ {
+  "pr": 8,
+  "index": 0,
+  "methods": [
+   "pkg/api/webassets/webassets.go::GetWebAssets"
+  ],
+  "file": "pkg/api/webassets/webassets.go",
+  "line": 48,
+  "why": "After acquiring the write lock it does not re-check entryPointAssetsCache, so goroutines that queued on the lock all re-read the assets."
+ },
+ {
+  "pr": 8,
+  "index": 1,
+  "methods": [
+   "pkg/api/webassets/webassets.go::GetWebAssets"
+  ],
+  "file": "pkg/api/webassets/webassets.go",
+  "line": 69,
+  "why": "entryPointAssetsCache = result runs even when the read failed, overwriting a valid cache with nil."
+ },
+ {
+  "pr": 9,
+  "index": 0,
+  "methods": [
+   "pkg/expr/reader.go::enableSqlExpressions"
+  ],
+  "file": "pkg/expr/reader.go",
+  "line": 195,
+  "why": "Both branches return false (lines 195-199), so ReadQuery always rejects SQL expressions."
+ },
+ {
+  "pr": 9,
+  "index": 1,
+  "methods": [
+   "pkg/expr/sql/db.go::DB.RunCommands",
+   "pkg/expr/sql/db.go::DB.QueryFramesInto",
+   "pkg/expr/sql/db.go::DB.TablesList"
+  ],
+  "file": "pkg/expr/sql/db.go",
+  "line": 17,
+  "why": "The replacement for go-duck is stubs that return 'not implemented' (lines 13, 17, 21), so TablesList and SQLCommand.Execute always fail."
+ },
+ {
+  "pr": 10,
+  "index": 0,
+  "methods": [
+   "pkg/storage/unified/search/bleve.go::bleveBackend.BuildIndex"
+  ],
+  "file": "pkg/storage/unified/search/bleve.go",
+  "line": 137,
+  "why": "cacheMu now guards only the final b.cache[key] assignment, so concurrent BuildIndex calls for the same key both build the index."
+ },
+ {
+  "pr": 10,
+  "index": 1,
+  "methods": [
+   "pkg/storage/unified/resource/search.go::searchSupport.init",
+   "pkg/storage/unified/search/bleve.go::bleveBackend.TotalDocs"
+  ],
+  "file": "pkg/storage/unified/resource/search.go",
+  "line": 216,
+  "why": "init calls s.search.TotalDocs() after starting the event watcher goroutine; TotalDocs (bleve.go line 146) ranges over b.cache without cacheMu while BuildIndex can write it."
+ }
+]

greptile/truth/keycloak.json

@@ -0,0 +1,297 @@
+[
+ {
+  "pr": 1,
+  "index": 0,
+  "methods": [
+   "services/src/main/java/org/keycloak/authentication/authenticators/browser/UsernameForm.java::UsernameForm.authenticate"
+  ],
+  "file": "services/src/main/java/org/keycloak/authentication/authenticators/browser/UsernameForm.java",
+  "line": 47,
+  "why": "Calls isConditionalPasskeysEnabled() with no argument, but the only definition (UsernamePasswordForm) takes a UserModel, so this does not compile."
+ },
+ {
+  "pr": 1,
+  "index": 1,
+  "methods": [
+   "services/src/main/java/org/keycloak/authentication/authenticators/browser/UsernamePasswordForm.java::UsernamePasswordForm.isConditionalPasskeysEnabled",
+   "services/src/main/java/org/keycloak/authentication/authenticators/browser/UsernamePasswordForm.java::UsernamePasswordForm.authenticate"
+  ],
+  "file": "services/src/main/java/org/keycloak/authentication/authenticators/browser/UsernamePasswordForm.java",
+  "line": 161,
+  "why": "The added `&& user != null` makes the check false on the first login page (no user yet), so authenticate() at line 115 never calls webauthnAuth.fillContextForm and conditional passkey UI is lost for ID-less login."
+ },
+ {
+  "pr": 2,
+  "index": 0,
+  "methods": [
+   "model/infinispan/src/main/java/org/keycloak/models/cache/infinispan/idp/InfinispanIdentityProviderStorageProvider.java::InfinispanIdentityProviderStorageProvider.getForLogin"
+  ],
+  "file": "model/infinispan/src/main/java/org/keycloak/models/cache/infinispan/idp/InfinispanIdentityProviderStorageProvider.java",
+  "line": 246,
+  "why": "Cached-id resolution calls session.identityProviders().getById(id), re-entering the caching provider, instead of idpDelegate.getById(id)."
+ },
+ {
+  "pr": 2,
+  "index": 1,
+  "methods": [],
+  "file": "testsuite/integration-arquillian/tests/base/src/test/java/org/keycloak/testsuite/organization/cache/OrganizationCacheTest.java",
+  "line": 381,
+  "why": "Test testCacheIDPForLogin (not in methods_read) registers cleanup for identityProviders().get(\"alias\") instead of \"idp-alias-\" + i, so the created IdPs are not removed (same mistake again at line 425 for idp-alias-20)."
+ },
+ {
+  "pr": 3,
+  "index": 0,
+  "methods": [
+   "authz/client/src/main/java/org/keycloak/authorization/client/util/crypto/AuthzClientCryptoProvider.java::AuthzClientCryptoProvider.getBouncyCastleProvider"
+  ],
+  "file": "authz/client/src/main/java/org/keycloak/authorization/client/util/crypto/AuthzClientCryptoProvider.java",
+  "line": 63,
+  "why": "getBouncyCastleProvider returns the provider of the default KeyStore type rather than a BouncyCastle provider."
+ },
+ {
+  "pr": 3,
+  "index": 1,
+  "methods": [
+   "authz/client/src/main/java/org/keycloak/authorization/client/util/crypto/AuthzClientCryptoProvider.java::AuthzClientCryptoProvider.getEcdsaCryptoProvider.concatenatedRSToASN1DER",
+   "authz/client/src/main/java/org/keycloak/authorization/client/util/crypto/AuthzClientCryptoProvider.java::AuthzClientCryptoProvider.getEcdsaCryptoProvider"
+  ],
+  "file": "authz/client/src/main/java/org/keycloak/authorization/client/util/crypto/AuthzClientCryptoProvider.java",
+  "line": 114,
+  "why": "Lines 114-115 create ASN1Encoder instances and write r and s, then discard them; the return statement re-encodes with fresh encoders (dead code inside the anonymous ECDSACryptoProvider)."
+ },
+ {
+  "pr": 3,
+  "index": 2,
+  "methods": [
+   "authz/client/src/main/java/org/keycloak/authorization/client/util/crypto/ASN1Decoder.java::ASN1Decoder.readSequence",
+   "authz/client/src/main/java/org/keycloak/authorization/client/util/crypto/ASN1Decoder.java::ASN1Decoder.readLength"
+  ],
+  "file": "authz/client/src/main/java/org/keycloak/authorization/client/util/crypto/ASN1Decoder.java",
+  "line": 60,
+  "why": "readSequence does `length = length - bytes.length` on the value from readLength, which returns -1 for indefinite-length encoding (line 134); the -1 is never handled, so such sequences are silently read as empty."
+ },
+ {
+  "pr": 3,
+  "index": 3,
+  "methods": [
+   "common/src/main/java/org/keycloak/common/crypto/CryptoProvider.java::CryptoProvider.order"
+  ],
+  "file": "common/src/main/java/org/keycloak/common/crypto/CryptoProvider.java",
+  "line": 44,
+  "why": "New abstract `int order()` on the public CryptoProvider interface (no default) breaks every existing third-party implementation."
+ },
+ {
+  "pr": 4,
+  "index": 0,
+  "methods": [
+   "quarkus/runtime/src/main/java/org/keycloak/quarkus/runtime/cli/command/UpdateCompatibilityCheck.java::UpdateCompatibilityCheck.run",
+   "quarkus/runtime/src/main/java/org/keycloak/quarkus/runtime/cli/command/UpdateCompatibilityMetadata.java::UpdateCompatibilityMetadata.run"
+  ],
+  "file": "quarkus/runtime/src/main/java/org/keycloak/quarkus/runtime/cli/command/UpdateCompatibilityCheck.java",
+  "line": 48,
+  "why": "Feature-disabled branch calls picocli.exit(CompatibilityResult.FEATURE_DISABLED), which calls System.exit directly; the identical new call is in UpdateCompatibilityMetadata.run line 48 (ambiguous which file the comment meant)."
+ },
+ {
+  "pr": 4,
+  "index": 1,
+  "methods": [],
+  "file": "quarkus/runtime/src/main/java/org/keycloak/quarkus/runtime/compatibility/CompatibilityResult.java",
+  "line": 34,
+  "why": "Constant declaration (not a method; covered only by the CompatibilityResult.java::<top-level> entry): RECREATE_UPGRADE_EXIT_CODE changes 4 to 3 and 4 becomes FEATURE_DISABLED with no release note or migration guidance; this is a missing-docs comment, so the location is the changed constant."
+ },
+ {
+  "pr": 5,
+  "index": 0,
+  "methods": [
+   "services/src/main/java/org/keycloak/services/resources/admin/permissions/AdminPermissions.java::AdminPermissions.registerListener.onEvent",
+   "services/src/main/java/org/keycloak/services/resources/admin/permissions/AdminPermissions.java::AdminPermissions.registerListener"
+  ],
+  "file": "services/src/main/java/org/keycloak/services/resources/admin/permissions/AdminPermissions.java",
+  "line": 77,
+  "why": "Removal-event cleanup is gated on Profile.Feature.ADMIN_FINE_GRAINED_AUTHZ (V1) instead of ADMIN_FINE_GRAINED_AUTHZ_V2, so with only V2 enabled permissions are orphaned."
+ },
+ {
+  "pr": 5,
+  "index": 1,
+  "methods": [
+   "services/src/main/java/org/keycloak/services/resources/admin/permissions/ClientPermissionsV2.java::ClientPermissionsV2.hasPermission",
+   "server-spi-private/src/main/java/org/keycloak/authorization/AdminPermissionsSchema.java::AdminPermissionsSchema.getOrCreateResource"
+  ],
+  "file": "services/src/main/java/org/keycloak/services/resources/admin/permissions/ClientPermissionsV2.java",
+  "line": 214,
+  "why": "findByName(server, client.getId(), server.getId()) uses the server id as owner, but getOrCreateResource creates per-client resources with owner resourceServer.getClientId() (line 104), so per-client resources are never found."
+ },
+ {
+  "pr": 5,
+  "index": 2,
+  "methods": [
+   "services/src/main/java/org/keycloak/services/resources/admin/permissions/ClientPermissionsV2.java::ClientPermissionsV2.getClientsWithPermission"
+  ],
+  "file": "services/src/main/java/org/keycloak/services/resources/admin/permissions/ClientPermissionsV2.java",
+  "line": 138,
+  "why": "findByType(CLIENTS_RESOURCE_TYPE) only yields the type-level 'Clients' resource (per-client resources have no type), and line 140 returns its name rather than client ids, which callers pass to realm.getClientById."
+ },
+ {
+  "pr": 5,
+  "index": 3,
+  "methods": [
+   "services/src/main/java/org/keycloak/services/resources/admin/permissions/ClientPermissionsV2.java::ClientPermissionsV2.hasPermission"
+  ],
+  "file": "services/src/main/java/org/keycloak/services/resources/admin/permissions/ClientPermissionsV2.java",
+  "line": 219,
+  "why": "getResourceTypeResource (line 217) can return null and the result is passed to findByResource and later resource.getScopes() without a null check."
+ },
+ {
+  "pr": 5,
+  "index": 4,
+  "methods": [
+   "services/src/main/java/org/keycloak/services/resources/admin/permissions/ClientPermissionsV2.java::ClientPermissionsV2.getEvaluationContext",
+   "services/src/main/java/org/keycloak/services/resources/admin/permissions/ClientPermissionsV2.java::ClientPermissionsV2.getEvaluationContext.getBaseAttributes"
+  ],
+  "file": "services/src/main/java/org/keycloak/services/resources/admin/permissions/ClientPermissionsV2.java",
+  "line": 262,
+  "why": "Private getEvaluationContext (with its anonymous getBaseAttributes override) is never called."
+ },
+ {
+  "pr": 6,
+  "index": 0,
+  "methods": [
+   "services/src/main/java/org/keycloak/services/resources/admin/permissions/GroupPermissionsV2.java::GroupPermissionsV2.canManage"
+  ],
+  "file": "services/src/main/java/org/keycloak/services/resources/admin/permissions/GroupPermissionsV2.java",
+  "line": 70,
+  "why": "canManage() falls back to hasPermission(null, VIEW, MANAGE), so a VIEW-only grant passes the manage check; it should check MANAGE only."
+ },
+ {
+  "pr": 6,
+  "index": 1,
+  "methods": [
+   "services/src/main/java/org/keycloak/services/resources/admin/permissions/GroupPermissionsV2.java::GroupPermissionsV2.getGroupIdsWithViewPermission",
+   "services/src/main/java/org/keycloak/services/resources/admin/permissions/GroupPermissionsV2.java::GroupPermissionsV2.hasPermission"
+  ],
+  "file": "services/src/main/java/org/keycloak/services/resources/admin/permissions/GroupPermissionsV2.java",
+  "line": 122,
+  "why": "Passes and collects groupResource.getId() (resource id) although hasPermission resolves resources by name (the group id) and callers expect group ids, so per-group grants do not yield the right ids."
+ },
+ {
+  "pr": 7,
+  "index": 0,
+  "methods": [],
+  "file": "themes/src/main/resources-community/theme/base/account/messages/messages_lt.properties",
+  "line": 101,
+  "why": "Translation file (no method): totpStep1 in the Lithuanian file is replaced with Italian text; the same Italian line is in login/messages/messages_lt.properties line 71 (loginTotpStep1), so the comment could mean either."
+ },
+ {
+  "pr": 7,
+  "index": 1,
+  "methods": [],
+  "file": "themes/src/main/resources-community/theme/base/account/messages/messages_zh_CN.properties",
+  "line": 112,
+  "why": "Translation file (no method): totpStep1 in the Simplified Chinese file uses Traditional Chinese characters."
+ },
+ {
+  "pr": 7,
+  "index": 2,
+  "methods": [
+   "misc/theme-verifier/src/main/java/org/keycloak/themeverifier/VerifyMessageProperties.java::VerifyMessageProperties.santizeAnchors"
+  ],
+  "file": "misc/theme-verifier/src/main/java/org/keycloak/themeverifier/VerifyMessageProperties.java",
+  "line": 154,
+  "why": "Each translated anchor consumes the next English anchor via englishMatcher.find() and compares in order, so extra, missing, or reordered anchors give wrong validation results."
+ },
+ {
+  "pr": 7,
+  "index": 3,
+  "methods": [
+   "misc/theme-verifier/src/main/java/org/keycloak/themeverifier/VerifyMessageProperties.java::VerifyMessageProperties.santizeAnchors"
+  ],
+  "file": "misc/theme-verifier/src/main/java/org/keycloak/themeverifier/VerifyMessageProperties.java",
+  "line": 150,
+  "why": "Method name is misspelled santizeAnchors (should be sanitizeAnchors)."
+ },
+ {
+  "pr": 7,
+  "index": 4,
+  "methods": [],
+  "file": "misc/theme-verifier/src/main/java/org/keycloak/themeverifier/VerifyMessageProperties.java",
+  "line": 139,
+  "why": "Field declaration (not a method; covered only by the VerifyMessageProperties.java::<top-level> entry): `Pattern HTML_TAGS` is package-private and non-final instead of private static final."
+ },
+ {
+  "pr": 8,
+  "index": 0,
+  "methods": [
+   "services/src/main/java/org/keycloak/protocol/oidc/encode/AccessTokenContext.java::AccessTokenContext.AccessTokenContext"
+  ],
+  "file": "services/src/main/java/org/keycloak/protocol/oidc/encode/AccessTokenContext.java",
+  "line": 73,
+  "why": "Fourth Objects.requireNonNull checks grantType again (with the rawTokenId message) instead of rawTokenId."
+ },
+ {
+  "pr": 8,
+  "index": 1,
+  "methods": [
+   "testsuite/integration-arquillian/tests/base/src/test/java/org/keycloak/testsuite/AssertEvents.java::AssertEvents.isAccessTokenId.matchesSafely",
+   "testsuite/integration-arquillian/tests/base/src/test/java/org/keycloak/testsuite/AssertEvents.java::AssertEvents.isAccessTokenId"
+  ],
+  "file": "testsuite/integration-arquillian/tests/base/src/test/java/org/keycloak/testsuite/AssertEvents.java",
+  "line": 483,
+  "why": "Uses substring(3,5) instead of substring(4,6) for the grant shortcut and returns false on a match, inverting the check."
+ },
+ {
+  "pr": 8,
+  "index": 2,
+  "methods": [
+   "server-spi-private/src/main/java/org/keycloak/protocol/oidc/grants/OAuth2GrantTypeFactory.java::OAuth2GrantTypeFactory.getShortcut"
+  ],
+  "file": "server-spi-private/src/main/java/org/keycloak/protocol/oidc/grants/OAuth2GrantTypeFactory.java",
+  "line": 30,
+  "why": "Javadoc on getShortcut() says 'usually like 3-letters shortcut' while the shortcuts and the 6-char encoding use 2 letters (comment sits just above the one-line method range 33)."
+ },
+ {
+  "pr": 8,
+  "index": 3,
+  "methods": [],
+  "file": "services/src/test/java/org/keycloak/protocol/oidc/encode/DefaultTokenContextEncoderProviderTest.java",
+  "line": 82,
+  "why": "Test method testIncorrectGrantType (not in methods_read) catches generic RuntimeException where the provider throws IllegalArgumentException."
+ },
+ {
+  "pr": 9,
+  "index": 0,
+  "methods": [
+   "services/src/main/java/org/keycloak/forms/login/freemarker/model/RecoveryAuthnCodeInputLoginBean.java::RecoveryAuthnCodeInputLoginBean.RecoveryAuthnCodeInputLoginBean"
+  ],
+  "file": "services/src/main/java/org/keycloak/forms/login/freemarker/model/RecoveryAuthnCodeInputLoginBean.java",
+  "line": 19,
+  "why": "Calls credentialModelOpt.get() on RecoveryAuthnCodesUtils.getCredential(user) without isPresent(), risking NoSuchElementException; the comment's raw-List title matches the greptile entry below instead (ambiguous comment, located by its body)."
+ },
+ {
+  "pr": 9,
+  "index": 1,
+  "methods": [
+   "testsuite/integration-arquillian/servers/auth-server/services/testsuite-providers/src/main/java/org/keycloak/testsuite/federation/BackwardsCompatibilityUserStorage.java::BackwardsCompatibilityUserStorage.getCredentials"
+  ],
+  "file": "testsuite/integration-arquillian/servers/auth-server/services/testsuite-providers/src/main/java/org/keycloak/testsuite/federation/BackwardsCompatibilityUserStorage.java",
+  "line": 237,
+  "why": "RecoveryAuthnCodesCredentialModel built via createFromValues never gets the stored credential's id (myUser.recoveryCodes.getId()), so getId() is null and removal by id fails."
+ },
+ {
+  "pr": 10,
+  "index": 0,
+  "methods": [
+   "model/infinispan/src/main/java/org/keycloak/models/cache/infinispan/GroupAdapter.java::GroupAdapter.getSubGroupsCount"
+  ],
+  "file": "model/infinispan/src/main/java/org/keycloak/models/cache/infinispan/GroupAdapter.java",
+  "line": 275,
+  "why": "Returns null when the group model is gone, violating GroupModel's documented never-null contract for getSubGroupsCount."
+ },
+ {
+  "pr": 10,
+  "index": 1,
+  "methods": [],
+  "file": "tests/base/src/test/java/org/keycloak/tests/admin/group/GroupTest.java",
+  "line": 155,
+  "why": "Test createMultiDeleteMultiReadMulti (not in methods_read) sets deletedAll and asserts at line 157 without joining the reader thread started at line 139, so late exceptions can be missed."
+ }
+]

greptile/truth/sentry.json

@@ -0,0 +1,349 @@
+[
+ {
+  "pr": 1,
+  "index": 0,
+  "methods": [
+   "src/sentry/api/paginator.py::OptimizedCursorPaginator.get_result"
+  ],
+  "file": "src/sentry/api/paginator.py",
+  "line": 882,
+  "why": "When enable_advanced_features is set and cursor.offset < 0, start_offset is the raw negative offset (line 880) and queryset[start_offset:stop] slices a Django QuerySet with a negative index, which Django rejects."
+ },
+ {
+  "pr": 1,
+  "index": 1,
+  "methods": [
+   "src/sentry/api/endpoints/organization_auditlogs.py::OrganizationAuditLogsEndpoint.get"
+  ],
+  "file": "src/sentry/api/endpoints/organization_auditlogs.py",
+  "line": 71,
+  "why": "enable_advanced reads organization_context.member.has_global_access with no None check; member is None for API-key/org-token requests, so the attribute access raises AttributeError (it is evaluated even when optimized_pagination is not requested, because it runs unconditionally for non-superusers)."
+ },
+ {
+  "pr": 1,
+  "index": 2,
+  "methods": [
+   "src/sentry/api/paginator.py::OptimizedCursorPaginator.get_item_key"
+  ],
+  "file": "src/sentry/api/paginator.py",
+  "line": 840,
+  "why": "get_item_key applies math.floor/math.ceil to getattr(item, self.key), but the audit log endpoint uses it with order_by='-datetime', so the key is a datetime and floor/ceil raise TypeError (DateTimePaginator converts to a timestamp first)."
+ },
+ {
+  "pr": 1,
+  "index": 3,
+  "methods": [],
+  "file": "src/sentry/api/endpoints/organization_auditlogs.py",
+  "line": 11,
+  "why": "Module-level import of OptimizedCursorPaginator (outside the organization_auditlogs.py <top-level> range 17-40, which excludes the import block); at the head commit the class does exist in paginator.py (line 821, added by this PR), so the 'non-existent' claim does not hold here."
+ },
+ {
+  "pr": 2,
+  "index": 0,
+  "methods": [
+   "src/sentry/api/paginator.py::OptimizedCursorPaginator.get_result"
+  ],
+  "file": "src/sentry/api/paginator.py",
+  "line": 880,
+  "why": "In the enable_advanced_features and cursor.offset < 0 branch, cursor.offset is used directly as start_offset and passed to queryset[start_offset:stop] (line 882) with no clamping."
+ },
+ {
+  "pr": 2,
+  "index": 1,
+  "methods": [
+   "src/sentry/api/paginator.py::BasePaginator.get_result"
+  ],
+  "file": "src/sentry/api/paginator.py",
+  "line": 182,
+  "why": "start_offset = max(0, offset) if not cursor.is_prev else offset leaves prev-cursor offsets unclamped, so a negative offset from a crafted cursor reaches queryset slicing on line 184."
+ },
+ {
+  "pr": 2,
+  "index": 2,
+  "methods": [
+   "src/sentry/api/paginator.py::OptimizedCursorPaginator.get_item_key"
+  ],
+  "file": "src/sentry/api/paginator.py",
+  "line": 840,
+  "why": "math.floor/math.ceil are applied to the raw datetime key (order_by='-datetime'), which raises TypeError."
+ },
+ {
+  "pr": 2,
+  "index": 3,
+  "methods": [
+   "src/sentry/api/endpoints/organization_auditlogs.py::OrganizationAuditLogsEndpoint.get"
+  ],
+  "file": "src/sentry/api/endpoints/organization_auditlogs.py",
+  "line": 71,
+  "why": "organization_context.member.has_global_access is read without checking that member is not None, which raises AttributeError for non-member / token-authenticated callers that are not superusers."
+ },
+ {
+  "pr": 3,
+  "index": 0,
+  "methods": [
+   "src/sentry/testutils/factories.py::_set_sample_rate_from_error_sampling"
+  ],
+  "file": "src/sentry/testutils/factories.py",
+  "line": 353,
+  "why": "`if client_sample_rate:` is falsy for 0.0, so a client_sample_rate of 0.0 never sets normalized_data['sample_rate']; it should test `is not None`."
+ },
+ {
+  "pr": 3,
+  "index": 1,
+  "methods": [
+   "src/sentry/api/helpers/error_upsampling.py::is_errors_query_for_error_upsampled_projects",
+   "src/sentry/api/helpers/error_upsampling.py::invalidate_upsampling_cache"
+  ],
+  "file": "src/sentry/api/helpers/error_upsampling.py",
+  "line": 27,
+  "why": "The cache key embeds hash(tuple(sorted(project_ids))) and invalidate_upsampling_cache rebuilds it the same way (line 73); hash() is not a stable serialization for a shared cache key. Note that CPython only randomizes str/bytes hashes, so for int project ids the keys do match across workers today; the location is right but the stated failure is overstated."
+ },
+ {
+  "pr": 3,
+  "index": 2,
+  "methods": [
+   "src/sentry/api/endpoints/organization_events_stats.py::OrganizationEventsStatsEndpoint.get._get_event_stats"
+  ],
+  "file": "src/sentry/api/endpoints/organization_events_stats.py",
+  "line": 221,
+  "why": "is_errors_query_for_error_upsampled_projects is passed the enclosing `dataset` instead of the `scoped_dataset` parameter, so when the dashboard split path calls _get_event_stats with discover (or errors) while the request dataset is metrics, the eligibility check uses the wrong dataset."
+ },
+ {
+  "pr": 4,
+  "index": 0,
+  "methods": [
+   "src/sentry/integrations/github/integration.py::GitHubInstallation.dispatch"
+  ],
+  "file": "src/sentry/integrations/github/integration.py",
+  "line": 502,
+  "why": "pipeline.fetch_state('github_authenticated_user') is used without checking it was set (e.g. when the OAuth step was skipped); at head a missing value compares None against the sender login and fails closed with the generic error rather than crashing, so this is the most defensible location."
+ },
+ {
+  "pr": 4,
+  "index": 1,
+  "methods": [
+   "src/sentry/integrations/github/integration.py::OAuthLoginView.dispatch"
+  ],
+  "file": "src/sentry/integrations/github/integration.py",
+  "line": 402,
+  "why": "The OAuth state parameter is pipeline.signature, a deterministic md5 of the pipeline ids, rather than a per-request random nonce; it is compared against the same value on line 412."
+ },
+ {
+  "pr": 4,
+  "index": 2,
+  "methods": [
+   "src/sentry/integrations/github/integration.py::GitHubInstallation.dispatch"
+  ],
+  "file": "src/sentry/integrations/github/integration.py",
+  "line": 503,
+  "why": "integration.metadata['sender']['login'] is indexed directly, raising KeyError for existing integrations whose metadata has no 'sender' entry."
+ },
+ {
+  "pr": 5,
+  "index": 0,
+  "methods": [
+   "src/sentry/issues/endpoints/browser_reporting_collector.py::BrowserReportingCollectorEndpoint.post"
+  ],
+  "file": "src/sentry/issues/endpoints/browser_reporting_collector.py",
+  "line": 91,
+  "why": "The comment describes the base code: this PR replaced the HttpResponse returns with rest_framework Response (lines 91, 104, 115, 130), so at head post already returns DRF Responses; the location is still this method."
+ },
+ {
+  "pr": 5,
+  "index": 1,
+  "methods": [
+   "src/sentry/workflow_engine/endpoints/validators/base/detector.py::BaseDetectorTypeValidator.update"
+  ],
+  "file": "src/sentry/workflow_engine/endpoints/validators/base/detector.py",
+  "line": 64,
+  "why": "instance.type is taken from validated_data.get('detector_type', ...), but the serializer field is named 'type', so a type change in an update is ignored and the existing group_type slug is kept."
+ },
+ {
+  "pr": 5,
+  "index": 2,
+  "methods": [
+   "src/sentry/replays/endpoints/project_replay_summarize_breadcrumbs.py::fetch_error_details"
+  ],
+  "file": "src/sentry/replays/endpoints/project_replay_summarize_breadcrumbs.py",
+  "line": 118,
+  "why": "zip(error_ids, events.values()) pairs ids with nodestore get_multi results by position, so a missing node or a different dict order attaches event data to the wrong error id; it should look up events[node_id] per id."
+ },
+ {
+  "pr": 5,
+  "index": 3,
+  "methods": [
+   "static/app/views/dashboards/widgetCard/chart.tsx::WidgetCardChart.tableResultComponent"
+  ],
+  "file": "static/app/views/dashboards/widgetCard/chart.tsx",
+  "line": 168,
+  "why": "When the use-table-widget-visualization flag is on, TableWidgetVisualization is rendered with columns={[]} and tableData data: [] and empty meta (lines 165-174) instead of result.data/result.meta, so the table is always empty."
+ },
+ {
+  "pr": 5,
+  "index": 4,
+  "methods": [
+   "src/sentry/issues/endpoints/browser_reporting_collector.py::BrowserReportSerializer.validate_age",
+   "src/sentry/issues/endpoints/browser_reporting_collector.py::BrowserReportSerializer.validate_timestamp"
+  ],
+  "file": "src/sentry/issues/endpoints/browser_reporting_collector.py",
+  "line": 58,
+  "why": "validate_age checks self.initial_data.get('timestamp'), which is falsy for timestamp=0 (allowed by min_value=0), so age and timestamp=0 pass together; validate_timestamp has the same truthiness check on 'age' (line 52). Both should use `in self.initial_data`."
+ },
+ {
+  "pr": 5,
+  "index": 5,
+  "methods": [
+   "src/sentry/preprod/api/endpoints/organization_preprod_artifact_assemble.py::ProjectPreprodArtifactAssembleEndpoint.post"
+  ],
+  "file": "src/sentry/preprod/api/endpoints/organization_preprod_artifact_assemble.py",
+  "line": 81,
+  "why": "analytics.record('preprod_artifact.api.assemble', ...) runs before the organizations:preprod-artifact-assemble feature check on line 88, so requests that get the 404 are still recorded."
+ },
+ {
+  "pr": 6,
+  "index": 0,
+  "methods": [
+   "src/sentry/spans/consumers/process/flusher.py::SpanFlusher.main"
+  ],
+  "file": "src/sentry/spans/consumers/process/flusher.py",
+  "line": 199,
+  "why": "spans.buffer.flusher.wait_produce is tagged {'shards': shard_tag} while flusher.produce (line 185) and segment_size_bytes (line 195) use {'shard': shard_tag}."
+ },
+ {
+  "pr": 6,
+  "index": 1,
+  "methods": [],
+  "file": "tests/sentry/spans/consumers/process/test_consumer.py",
+  "line": 62,
+  "why": "In test_basic (test file not listed in methods_read), time.sleep is monkeypatched to a no-op on line 15, so time.sleep(0.1) on line 62 does not wait for the flusher."
+ },
+ {
+  "pr": 6,
+  "index": 2,
+  "methods": [
+   "src/sentry/spans/consumers/process/flusher.py::SpanFlusher._ensure_processes_alive"
+  ],
+  "file": "src/sentry/spans/consumers/process/flusher.py",
+  "line": 254,
+  "why": "Processes come from mp_context = multiprocessing.get_context('spawn').Process, i.e. SpawnProcess, which is not a subclass of multiprocessing.Process, so isinstance(process, multiprocessing.Process) is False and hung processes are never killed before a new one is created."
+ },
+ {
+  "pr": 6,
+  "index": 3,
+  "methods": [],
+  "file": "tests/sentry/spans/consumers/process/test_consumer.py",
+  "line": 62,
+  "why": "Same as golden 1: test_basic's time.sleep(0.1) is a no-op because of the monkeypatch on line 15 (test file not listed in methods_read)."
+ },
+ {
+  "pr": 6,
+  "index": 4,
+  "methods": [
+   "src/sentry/spans/consumers/process/flusher.py::SpanFlusher.join"
+  ],
+  "file": "src/sentry/spans/consumers/process/flusher.py",
+  "line": 341,
+  "why": "Once the deadline has passed, `break` exits the loop before the terminate() call on line 347, so the remaining flusher processes are never terminated."
+ },
+ {
+  "pr": 7,
+  "index": 0,
+  "methods": [
+   "src/sentry/integrations/services/assignment_source.py::<top-level>"
+  ],
+  "file": "src/sentry/integrations/services/assignment_source.py",
+  "line": 18,
+  "why": "In the AssignmentSource class body, `queued: datetime = timezone.now()` is evaluated once at import, so every instance shares the module-load timestamp; it needs field(default_factory=timezone.now)."
+ },
+ {
+  "pr": 7,
+  "index": 1,
+  "methods": [],
+  "file": "tests/sentry/integrations/services/test_assignment_source.py",
+  "line": 13,
+  "why": "Test method name test_from_dict_inalid_data is misspelled (test file not listed in methods_read)."
+ },
+ {
+  "pr": 7,
+  "index": 2,
+  "methods": [],
+  "file": "tests/sentry/integrations/services/test_assignment_source.py",
+  "line": 8,
+  "why": "test_from_dict_empty_array passes an empty dict, so the name is misleading (test file not listed in methods_read)."
+ },
+ {
+  "pr": 7,
+  "index": 3,
+  "methods": [
+   "src/sentry/integrations/services/assignment_source.py::AssignmentSource.to_dict",
+   "src/sentry/integrations/utils/sync.py::sync_group_assignee_outbound"
+  ],
+  "file": "src/sentry/integrations/services/assignment_source.py",
+  "line": 28,
+  "why": "to_dict returns asdict(self), which keeps queued as a datetime; sync_group_assignee_outbound puts that dict in apply_async kwargs (sync.py line 141), where a JSON task serializer cannot encode it."
+ },
+ {
+  "pr": 8,
+  "index": 0,
+  "methods": [
+   "src/sentry/monitors/logic/incident_occurrence.py::get_monitor_environment_context"
+  ],
+  "file": "src/sentry/monitors/logic/incident_occurrence.py",
+  "line": 168,
+  "why": "config is copied and schedule_type replaced with its display value (lines 160-162), but the returned dict uses monitor_environment.monitor.config, discarding the modified copy."
+ },
+ {
+  "pr": 8,
+  "index": 1,
+  "methods": [
+   "src/sentry/monitors/logic/incidents.py::try_incident_threshold"
+  ],
+  "file": "src/sentry/monitors/logic/incidents.py",
+  "line": 93,
+  "why": "MonitorCheckIn.objects.filter(id__in=[c['id'] for c in previous_checkins]) re-queries check-ins whose data is already in previous_checkins."
+ },
+ {
+  "pr": 9,
+  "index": 0,
+  "methods": [],
+  "file": "tests/sentry/uptime/consumers/test_results_consumer.py",
+  "line": 1718,
+  "why": "max_wait = 50 is a repeated literal in three thread-queue tests (lines 1718, 1765, 1812; other tests use 20); a style comment, located at the first occurrence."
+ },
+ {
+  "pr": 9,
+  "index": 1,
+  "methods": [],
+  "file": "tests/sentry/uptime/consumers/test_results_consumer.py",
+  "line": 1876,
+  "why": "The docstring says errors do not block offset commits for other messages, but the test asserts that nothing was committed for the partition (line 1922)."
+ },
+ {
+  "pr": 10,
+  "index": 0,
+  "methods": [
+   "src/sentry/incidents/grouptype.py::<top-level>"
+  ],
+  "file": "src/sentry/incidents/grouptype.py",
+  "line": 11,
+  "why": "MetricAlertDetectorHandler subclasses StatefulDetectorHandler with only `pass`, leaving counter_names, get_dedupe_value, get_group_key_values and build_occurrence_and_event_data abstract, so instantiating it raises TypeError; class-body code, so the file's <top-level>."
+ },
+ {
+  "pr": 10,
+  "index": 1,
+  "methods": [
+   "src/sentry/workflow_engine/processors/detector.py::StatefulDetectorHandler.evaluate"
+  ],
+  "file": "src/sentry/workflow_engine/processors/detector.py",
+  "line": 230,
+  "why": "The docstring still says evaluate returns a list of DetectorEvaluationResult, but it now returns a dict keyed by DetectorGroupKey."
+ },
+ {
+  "pr": 10,
+  "index": 2,
+  "methods": [],
+  "file": "tests/sentry/workflow_engine/processors/test_detector.py",
+  "line": 191,
+  "why": "The expected occurrence for group_2 is built with value 6, but the data packet on line 177 gives group_2 the value 10."
+ }
+]

reviewbench/README.md

@@ -0,0 +1,26 @@
+# ReviewBench
+
+Perch's runs on [ReviewBench](https://github.com/review-bench/ReviewBench), GitHub's benchmark for AI code review. The scripts that score these files are in [`.benchmarks/reviewbench`](https://github.com/lakeday-org/perch/tree/main/.benchmarks/reviewbench) in lakeday-org/perch. Its README describes both runs and how they are scored.
+
+These files are scored by lakeday-org/perch at commit `6b33c162a778f56e27ae043f57f4d52cf30b9456`, from [lakeday-org/perch#421](https://github.com/lakeday-org/perch/pull/421). Download them into a perch checkout, where the scripts read them:
+
+```sh
+hf download perchscan/benchmark-results --repo-type dataset --include 'reviewbench/*' --local-dir .benchmarks/hf
+```
+
+| File | Contents |
+|---|---|
+| `targets.jsonl` | The 16,168 methods scored, each with its repository, commits, file, lines, split and labels |
+| `requests.jsonl.gz` | The frozen run's requests, two readings of each method |
+| `results/jev.jsonl.gz` | The frozen run's answers from Jev 1.14 |
+| `results/luna.jsonl.gz` | The frozen run's answers from GPT-6 Luna Decisions |
+| `judgments.jsonl` | Judgments of the alerts ReviewBench did not label, with quoted source |
+| `coverage.jsonl` | How many methods were extracted from each file of all 219 pull requests |
+| `manifest.json` | The ReviewBench and perch revisions used, and each file's SHA-256 hash |
+| `summary.json` | The frozen run's scores, from `score.mjs` |
+| `results/rerun-luna.jsonl.gz` | The rerun's answers, one line a pull request, from `perch scan --since` on GPT-6 Luna Decisions |
+| `rerun-summary.json` | The rerun's scores, from `rerun-score.mjs` |
+| `results/rerun-jev.jsonl.gz` | The same rerun on Jev (`jev-latest`), from the released perch 0.4.8 |
+| `rerun-jev-summary.json` | The Jev rerun's scores, from `rerun-score.mjs --results` |
+
+Source code quoted in these files is licensed according to its repository. ReviewBench is [MIT-licensed](https://github.com/review-bench/ReviewBench/blob/main/LICENSE).

reviewbench/manifest.json

@@ -0,0 +1,59 @@
+{
+  "schema_version": 1,
+  "benchmark": "ReviewBench",
+  "upstream": "https://github.com/review-bench/ReviewBench",
+  "upstream_revision": "c17b05be9c90b85a1fc012ce39e455e2d09a7255",
+  "perch_revision": "67575f846d2b0c1d912fea8e0f0e9ba8941c4a24",
+  "selected_prs": 219,
+  "prs_with_targets": 177,
+  "targets": 16168,
+  "positive_targets": 1506,
+  "failure_policy": "A failed target is a miss if positive and a false alert if clean. Missing requests are incomplete coverage, never scored as clean.",
+  "thresholds": {
+    "functional": "strictly greater than 0.60",
+    "security": "strictly greater than 0.70"
+  },
+  "models": {
+    "jev": "jev-1.14.0",
+    "luna": "gpt-6-luna"
+  },
+  "input_prices_per_million": {
+    "jev": 0.042,
+    "luna": 0.1
+  },
+  "sha256": {
+    "targets.jsonl": "476f46f98e1ca72e8e17e8b65e1867d7a45b94140b72b1d7489c348634164adf",
+    "requests.jsonl.gz": "bb38d8ec68c16004e04a46ff489d072bc5bb036cfc795d53ae5271c34328a05e",
+    "judgments.jsonl": "09add93765a34161597e981da463d30c8998f0e7497a697e38ecd8f8a11be5db",
+    "coverage.jsonl": "a8a33d5fc42d6b5cd4805bcad8091b06153e7ac8d6101e667a472bec3bf530b0",
+    "results/jev.jsonl.gz": "f946f88bdf850f165b0541ca9247c3d3f1a856bcf8666bd20de455709d0d66d7",
+    "results/luna.jsonl.gz": "219d746feaaa2b58ac8a7750913542f9f017123dcaf97e19463ed18252a2bcd7"
+  },
+  "protocol": {
+    "readings": [
+      "change_current",
+      "relation"
+    ],
+    "selection": "The same fixed two-reading protocol for both providers."
+  },
+  "readings_per_model": 30936,
+  "planned_api_calls_per_model": 30945,
+  "completed_at": "2026-10-15T21:03:15.289Z",
+  "graph_revision": "eaf1cc93313cc3329b06e551f0b42922fc7a5f65",
+  "runs": {
+    "jev": {
+      "readings": 30936,
+      "api_calls_attempted": 30945,
+      "failed_readings": 0,
+      "failed_targets": 0
+    },
+    "luna": {
+      "readings": 30936,
+      "api_calls_attempted": 30945,
+      "failed_readings": 1,
+      "failed_targets": 1
+    }
+  },
+  "target_scope": "method",
+  "prs_without_methods": 42
+}

reviewbench/requests.jsonl.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bb38d8ec68c16004e04a46ff489d072bc5bb036cfc795d53ae5271c34328a05e
+size 12270547

reviewbench/rerun-jev-summary.json

@@ -0,0 +1,236 @@
+{
+ "targets": 16168,
+ "buggy": 1506,
+ "buggy_reference": 861,
+ "prs": 177,
+ "scanned_prs": 171,
+ "failed": [
+  {
+   "pr": "hughdidit_DAISy-Agency_274-1c88a238",
+   "code": "checkout",
+   "error": "Command failed: git clone -q --filter=blob:none --no-checkout https://github.com/hughdidit/DAISy-Agency.git /Users/jfbrown/code/perch-research/.claude/worktrees/greptile-bench/reviewbench/repos/hughdidit__DAISy-Agency"
+  },
+  {
+   "pr": "Nixbug_entebus-android-operator_67-21b50bbb",
+   "code": "checkout",
+   "error": "Command failed: git clone -q --filter=blob:none --no-checkout https://github.com/Nixbug/entebus-android-operator.git /Users/jfbrown/code/perch-research/.claude/worktrees/greptile-bench/reviewbench/repos/Nixbug__entebus-android-operator"
+  },
+  {
+   "pr": "hughdidit_DAISy-Agency_309-ac6e2196",
+   "code": "checkout",
+   "error": "Command failed: git clone -q --filter=blob:none --no-checkout https://github.com/hughdidit/DAISy-Agency.git /Users/jfbrown/code/perch-research/.claude/worktrees/greptile-bench/reviewbench/repos/hughdidit__DAISy-Agency"
+  },
+  {
+   "pr": "RobertAgterhuis_myAgentic-IT-Project-team-V2_331-980b84fa",
+   "code": "checkout",
+   "error": "Command failed: git clone -q --filter=blob:none --no-checkout https://github.com/RobertAgterhuis/myAgentic-IT-Project-team-V2.git /Users/jfbrown/code/perch-research/.claude/worktrees/greptile-bench/reviewbench/repos/RobertAgterhuis__myAgentic-IT-Project-team-V2"
+  },
+  {
+   "pr": "anokye-labs_kbexplorer_38-cf24fe27",
+   "code": "checkout",
+   "error": "base 84bf940b71ac021cbe2afa4b1f445959774c4d63 not fetchable"
+  },
+  {
+   "pr": "whispr-messenger_mobile-app_64-9e7742e5",
+   "code": "checkout",
+   "error": "Command failed: git clone -q --filter=blob:none --no-checkout https://github.com/whispr-messenger/mobile-app.git /Users/jfbrown/code/perch-research/.claude/worktrees/greptile-bench/reviewbench/repos/whispr-messenger__mobile-app"
+  }
+ ],
+ "readings": 17250,
+ "targets_read": 12955,
+ "unread_buggy": {
+  "in_failed_prs": 80,
+  "in_unparsed_files": 1,
+  "in_unread_functions": 292
+ },
+ "usage": {
+  "jev-1.14.0": {
+   "requests": 17369,
+   "input": 42814227,
+   "output": 2555058,
+   "cost": 1.798197534
+  }
+ },
+ "cost_usd": 1.798197534,
+ "cutoffs": [
+  {
+   "cutoff": 0.6,
+   "total": {
+    "targets": 16168,
+    "buggy": 1506,
+    "flagged": 260,
+    "flag_rate": 0.016081147946561108,
+    "precision": 0.4576923076923077,
+    "recall": 0.07901726427622842,
+    "f1": 0.1347678369195923,
+    "correct": 119,
+    "false": 141,
+    "missed": 1387
+   },
+   "development": {
+    "targets": 10326,
+    "buggy": 944,
+    "flagged": 207,
+    "flag_rate": 0.020046484601975594,
+    "precision": 0.47342995169082125,
+    "recall": 0.1038135593220339,
+    "f1": 0.17028670721112077,
+    "correct": 98,
+    "false": 109,
+    "missed": 846
+   },
+   "validation": {
+    "targets": 5842,
+    "buggy": 562,
+    "flagged": 53,
+    "flag_rate": 0.00907223553577542,
+    "precision": 0.39622641509433965,
+    "recall": 0.037366548042704624,
+    "f1": 0.06829268292682927,
+    "correct": 21,
+    "false": 32,
+    "missed": 541
+   },
+   "reference_only": {
+    "targets": 16168,
+    "buggy": 861,
+    "flagged": 260,
+    "flag_rate": 0.016081147946561108,
+    "precision": 0.3038461538461538,
+    "recall": 0.09175377468060394,
+    "f1": 0.1409455842997324,
+    "correct": 79,
+    "false": 181,
+    "missed": 782
+   }
+  },
+  {
+   "cutoff": 0.7,
+   "total": {
+    "targets": 16168,
+    "buggy": 1506,
+    "flagged": 81,
+    "flag_rate": 0.0050098960910440376,
+    "precision": 0.5061728395061729,
+    "recall": 0.027224435590969456,
+    "f1": 0.0516698172652804,
+    "correct": 41,
+    "false": 40,
+    "missed": 1465
+   },
+   "development": {
+    "targets": 10326,
+    "buggy": 944,
+    "flagged": 68,
+    "flag_rate": 0.0065853186132093745,
+    "precision": 0.5441176470588235,
+    "recall": 0.03919491525423729,
+    "f1": 0.07312252964426877,
+    "correct": 37,
+    "false": 31,
+    "missed": 907
+   },
+   "validation": {
+    "targets": 5842,
+    "buggy": 562,
+    "flagged": 13,
+    "flag_rate": 0.0022252653200958575,
+    "precision": 0.3076923076923077,
+    "recall": 0.0071174377224199285,
+    "f1": 0.01391304347826087,
+    "correct": 4,
+    "false": 9,
+    "missed": 558
+   },
+   "reference_only": {
+    "targets": 16168,
+    "buggy": 861,
+    "flagged": 81,
+    "flag_rate": 0.0050098960910440376,
+    "precision": 0.3333333333333333,
+    "recall": 0.0313588850174216,
+    "f1": 0.05732484076433121,
+    "correct": 27,
+    "false": 54,
+    "missed": 834
+   }
+  },
+  {
+   "cutoff": 0.8,
+   "total": {
+    "targets": 16168,
+    "buggy": 1506,
+    "flagged": 7,
+    "flag_rate": 0.0004329539831766452,
+    "precision": 0.8571428571428571,
+    "recall": 0.00398406374501992,
+    "f1": 0.007931262392597489,
+    "correct": 6,
+    "false": 1,
+    "missed": 1500
+   },
+   "development": {
+    "targets": 10326,
+    "buggy": 944,
+    "flagged": 5,
+    "flag_rate": 0.000484214603912454,
+    "precision": 1,
+    "recall": 0.005296610169491525,
+    "f1": 0.01053740779768177,
+    "correct": 5,
+    "false": 0,
+    "missed": 939
+   },
+   "validation": {
+    "targets": 5842,
+    "buggy": 562,
+    "flagged": 2,
+    "flag_rate": 0.00034234851078397807,
+    "precision": 0.5,
+    "recall": 0.0017793594306049821,
+    "f1": 0.0035460992907801418,
+    "correct": 1,
+    "false": 1,
+    "missed": 561
+   },
+   "reference_only": {
+    "targets": 16168,
+    "buggy": 861,
+    "flagged": 7,
+    "flag_rate": 0.0004329539831766452,
+    "precision": 0.5714285714285714,
+    "recall": 0.004645760743321719,
+    "f1": 0.009216589861751152,
+    "correct": 4,
+    "false": 3,
+    "missed": 857
+   }
+  }
+ ],
+ "frozen_reference_only": {
+  "jev": {
+   "targets": 16168,
+   "buggy": 861,
+   "flagged": 876,
+   "flag_rate": 0.05418109846610589,
+   "precision": 0.1780821917808219,
+   "recall": 0.18118466898954705,
+   "f1": 0.17962003454231434,
+   "correct": 156,
+   "false": 720,
+   "missed": 705
+  },
+  "luna": {
+   "targets": 16168,
+   "buggy": 861,
+   "flagged": 5635,
+   "flag_rate": 0.3485279564571994,
+   "precision": 0.10629991126885537,
+   "recall": 0.6957026713124274,
+   "f1": 0.18442118226600984,
+   "correct": 599,
+   "false": 5036,
+   "missed": 262
+  }
+ }
+}

reviewbench/rerun-summary.json

@@ -0,0 +1,236 @@
+{
+ "targets": 16168,
+ "buggy": 1506,
+ "buggy_reference": 861,
+ "prs": 177,
+ "scanned_prs": 171,
+ "failed": [
+  {
+   "pr": "hughdidit_DAISy-Agency_274-1c88a238",
+   "code": "checkout",
+   "error": "clone of https://github.com/hughdidit/DAISy-Agency failed"
+  },
+  {
+   "pr": "Nixbug_entebus-android-operator_67-21b50bbb",
+   "code": "checkout",
+   "error": "clone of https://github.com/Nixbug/entebus-android-operator failed"
+  },
+  {
+   "pr": "hughdidit_DAISy-Agency_309-ac6e2196",
+   "code": "checkout",
+   "error": "clone of https://github.com/hughdidit/DAISy-Agency failed"
+  },
+  {
+   "pr": "RobertAgterhuis_myAgentic-IT-Project-team-V2_331-980b84fa",
+   "code": "checkout",
+   "error": "clone of https://github.com/RobertAgterhuis/myAgentic-IT-Project-team-V2 failed"
+  },
+  {
+   "pr": "anokye-labs_kbexplorer_38-cf24fe27",
+   "code": "checkout",
+   "error": "base 84bf940b71ac021cbe2afa4b1f445959774c4d63 not fetchable"
+  },
+  {
+   "pr": "whispr-messenger_mobile-app_64-9e7742e5",
+   "code": "checkout",
+   "error": "clone of https://github.com/whispr-messenger/mobile-app failed"
+  }
+ ],
+ "readings": 17250,
+ "targets_read": 12955,
+ "unread_buggy": {
+  "in_failed_prs": 80,
+  "in_unparsed_files": 1,
+  "in_unread_functions": 292
+ },
+ "usage": {
+  "gpt-6-luna": {
+   "requests": 17369,
+   "input": 41569033,
+   "output": 0,
+   "cost": 4.1569033
+  }
+ },
+ "cost_usd": 4.1569033,
+ "cutoffs": [
+  {
+   "cutoff": 0.6,
+   "total": {
+    "targets": 16168,
+    "buggy": 1506,
+    "flagged": 2094,
+    "flag_rate": 0.12951509153884216,
+    "precision": 0.28796561604584525,
+    "recall": 0.40039840637450197,
+    "f1": 0.335,
+    "correct": 603,
+    "false": 1491,
+    "missed": 903
+   },
+   "development": {
+    "targets": 10326,
+    "buggy": 944,
+    "flagged": 1558,
+    "flag_rate": 0.15088127057912068,
+    "precision": 0.2920410783055199,
+    "recall": 0.4819915254237288,
+    "f1": 0.36370903277378097,
+    "correct": 455,
+    "false": 1103,
+    "missed": 489
+   },
+   "validation": {
+    "targets": 5842,
+    "buggy": 562,
+    "flagged": 536,
+    "flag_rate": 0.09174940089010612,
+    "precision": 0.27611940298507465,
+    "recall": 0.26334519572953735,
+    "f1": 0.26958105646630237,
+    "correct": 148,
+    "false": 388,
+    "missed": 414
+   },
+   "reference_only": {
+    "targets": 16168,
+    "buggy": 861,
+    "flagged": 2094,
+    "flag_rate": 0.12951509153884216,
+    "precision": 0.20152817574021012,
+    "recall": 0.4901277584204414,
+    "f1": 0.2856175972927242,
+    "correct": 422,
+    "false": 1672,
+    "missed": 439
+   }
+  },
+  {
+   "cutoff": 0.7,
+   "total": {
+    "targets": 16168,
+    "buggy": 1506,
+    "flagged": 1364,
+    "flag_rate": 0.08436417615042058,
+    "precision": 0.3519061583577713,
+    "recall": 0.3187250996015936,
+    "f1": 0.3344947735191638,
+    "correct": 480,
+    "false": 884,
+    "missed": 1026
+   },
+   "development": {
+    "targets": 10326,
+    "buggy": 944,
+    "flagged": 1007,
+    "flag_rate": 0.09752082122796824,
+    "precision": 0.3624627606752731,
+    "recall": 0.3866525423728814,
+    "f1": 0.3741670937980523,
+    "correct": 365,
+    "false": 642,
+    "missed": 579
+   },
+   "validation": {
+    "targets": 5842,
+    "buggy": 562,
+    "flagged": 357,
+    "flag_rate": 0.06110920917494009,
+    "precision": 0.32212885154061627,
+    "recall": 0.20462633451957296,
+    "f1": 0.250272034820457,
+    "correct": 115,
+    "false": 242,
+    "missed": 447
+   },
+   "reference_only": {
+    "targets": 16168,
+    "buggy": 861,
+    "flagged": 1364,
+    "flag_rate": 0.08436417615042058,
+    "precision": 0.25,
+    "recall": 0.39605110336817656,
+    "f1": 0.30651685393258427,
+    "correct": 341,
+    "false": 1023,
+    "missed": 520
+   }
+  },
+  {
+   "cutoff": 0.8,
+   "total": {
+    "targets": 16168,
+    "buggy": 1506,
+    "flagged": 836,
+    "flag_rate": 0.051707075705096486,
+    "precision": 0.41746411483253587,
+    "recall": 0.23173970783532535,
+    "f1": 0.29803586678052946,
+    "correct": 349,
+    "false": 487,
+    "missed": 1157
+   },
+   "development": {
+    "targets": 10326,
+    "buggy": 944,
+    "flagged": 611,
+    "flag_rate": 0.059171024598101876,
+    "precision": 0.4402618657937807,
+    "recall": 0.2849576271186441,
+    "f1": 0.3459807073954984,
+    "correct": 269,
+    "false": 342,
+    "missed": 675
+   },
+   "validation": {
+    "targets": 5842,
+    "buggy": 562,
+    "flagged": 225,
+    "flag_rate": 0.03851420746319754,
+    "precision": 0.35555555555555557,
+    "recall": 0.1423487544483986,
+    "f1": 0.20330368487928843,
+    "correct": 80,
+    "false": 145,
+    "missed": 482
+   },
+   "reference_only": {
+    "targets": 16168,
+    "buggy": 861,
+    "flagged": 836,
+    "flag_rate": 0.051707075705096486,
+    "precision": 0.30023923444976075,
+    "recall": 0.29152148664343785,
+    "f1": 0.2958161461402475,
+    "correct": 251,
+    "false": 585,
+    "missed": 610
+   }
+  }
+ ],
+ "frozen_reference_only": {
+  "jev": {
+   "targets": 16168,
+   "buggy": 861,
+   "flagged": 876,
+   "flag_rate": 0.05418109846610589,
+   "precision": 0.1780821917808219,
+   "recall": 0.18118466898954705,
+   "f1": 0.17962003454231434,
+   "correct": 156,
+   "false": 720,
+   "missed": 705
+  },
+  "luna": {
+   "targets": 16168,
+   "buggy": 861,
+   "flagged": 5635,
+   "flag_rate": 0.3485279564571994,
+   "precision": 0.10629991126885537,
+   "recall": 0.6957026713124274,
+   "f1": 0.18442118226600984,
+   "correct": 599,
+   "false": 5036,
+   "missed": 262
+  }
+ }
+}

reviewbench/results/jev.jsonl.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f946f88bdf850f165b0541ca9247c3d3f1a856bcf8666bd20de455709d0d66d7
+size 997667

reviewbench/results/luna.jsonl.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:219d746feaaa2b58ac8a7750913542f9f017123dcaf97e19463ed18252a2bcd7
+size 927113

reviewbench/results/rerun-jev.jsonl.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:781ed9caee71fef17f0b97bc1df2ec2837aeb1789fb6da9f1f457bbf7d4919ae
+size 553573

reviewbench/results/rerun-luna.jsonl.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:95f88bdd28969acfdd13f9a33333025a5c351877955fa0e8f1251d0ea76eb92a
+size 543282

reviewbench/summary.json

@@ -0,0 +1,28 @@
+[
+  {
+    "provider": "jev",
+    "targets": 16168,
+    "tp": 304,
+    "fp": 572,
+    "fn": 1202,
+    "tn": 14090,
+    "precision": 0.3470319634703196,
+    "recall": 0.20185922974767595,
+    "f1": 0.25524769101595296,
+    "failed_targets": 0,
+    "cost_usd": 2.928322194000014
+  },
+  {
+    "provider": "luna",
+    "targets": 16168,
+    "tp": 1221,
+    "fp": 4414,
+    "fn": 285,
+    "tn": 10248,
+    "precision": 0.21668145519077195,
+    "recall": 0.8107569721115537,
+    "f1": 0.34196891191709844,
+    "failed_targets": 1,
+    "cost_usd": 5.636321799999951
+  }
+]