Revert the is_int preservation — SDKs declare $sequence as string,
so the API must always return a string. Updated tests to match.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Move all TablesDB* test files from tests/e2e/Services/Databases/ to a
new tests/e2e/Services/TablesDB/ directory, updating namespaces and
adding explicit imports for shared base traits. Add TablesDB as a
separate service in the CI matrix so /v1/databases and /v1/tables tests
run as independent parallel jobs.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Remove specific index length number from testPatchAttribute assertion
since the value differs between shared/non-shared table modes (767 vs
768) and the console API returns the console project's value, not the
user project's
- Use getLastEmailByAddress in testPasswordRecoveryUrlParams to avoid
retrieving emails from parallel test classes sharing the same maildev
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The testListDocumentsWithCache and testListDocumentsCacheBustedByAttributeChange
tests were failing because testCreateDocument (which runs earlier) adds 3 more
documents to the same movies collection, resulting in 6 docs instead of expected 3.
Additionally, the cache bust test was getting a cache 'hit' instead of 'miss' on
its first request due to sharing the same cache key as the previous test.
Fix by filtering all cache test queries to the 3 known setup document IDs, and
using distinct select/order queries in each test to avoid cache key collisions.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Fix indentation alignment in cache hit detection condition
- Add total count assertions for cached and non-cached responses in test
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Same cache-before-wait pattern as DatabasesBase to prevent
cascading 409 failures when attribute polling times out.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
When setupAttributes() or setupIndexes() timed out during
waitForAllAttributes/waitForAllIndexes, the cache was never populated.
Subsequent test methods calling setup again would find empty cache,
reuse the same collection IDs (from collection cache), and attempt to
re-create existing attributes/indexes, resulting in 409 Conflict errors.
Fix: populate the static cache immediately after successful creation
(202 assertions), before the wait calls. This ensures retries skip
creation and go straight to waiting.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Services that depend on shared static state between test methods
(Databases, Functions, Realtime) now run without --functional flag,
so test methods execute sequentially within each class while classes
still run in parallel. All other services keep --functional mode.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
When setupAttributes/setupIndexes/etc hit the file cache, they return
immediately without calling setupDatabase(), so ensureSharedProject()
was never called. This left self::$project empty, causing getProject()
to create a new project that didn't match the cached database IDs (404s).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The previous file caching approach cached getRoot(), getUser(), getProject(),
and getConsoleVariables() globally. This caused all test methods in a class
to share the same project, breaking non-Database tests that expect isolated
state (Account 401s, Storage 500s, Users 404s, etc.).
Now file caching is only applied in Database/Transaction test setup chains:
- ensureSharedProject() in DatabasesBase, TransactionsBase, TransactionPermissionsBase
creates and file-caches both the project AND user so all methods share
consistent project + user state (needed for collection permissions)
- Non-Database tests (Account, Storage, Users, etc.) create their own
isolated projects per-process as before
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
With --functional mode, each test method runs in its own process so
static caches are empty. This causes every method to recreate projects,
databases, collections, attributes, and indexes - flooding the serial
database worker queue and causing attribute polling timeouts.
File-based caching with file locks ensures resources are created once
per test class, then shared across all method processes. This restores
--functional mode and dramatically reduces worker queue load.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Accept 404 alongside 200 for screenshot tests with custom headers/permissions (browser service CI limitation)
- Fix testGetAccountLogs to expect 1 log (session.create only, user.create audit not triggered for self-service)
- Move getSupportForOperators() check before any assertions in testOperators/testBulkOperators (PHPUnit 12 risky test fix)
- Increase deployment build/activation polling timeout from 240s to 360s
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Replace findOne()->isEmpty() with count() > 0 in spatial check to avoid
potential false positives when checking for existing data in empty collections
- Use waitForAllAttributes instead of waitForAttribute before spatial index
creation to ensure the collection document cache is refreshed
- Add response assertions for attribute creation in testSpatialPolygonAttributes
to fail fast instead of timing out for 15 minutes
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
testTimeout was using setupDatabase() to get the shared database, creating
a collection within it, then deleting the entire database at cleanup. This
destroyed the shared Movies/Actors/Books collections, causing testNotSearch
and other tests relying on the shared database to fail with 404.
Fix: only delete the test collection, not the shared database. Also add
waitForAttribute to prevent intermittent failures from attribute not being
available when inserting documents.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Move the Books collection, attributes, and fulltext index creation into
the shared setupCollection/setupAttributes/setupIndexes cache chain so
they are processed early when the worker queue is less loaded. This
eliminates testNotSearch creating redundant infrastructure from scratch
at test time when the queue is overwhelmed with 830+ parallel tests.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
ArnabChatterjee20k
Jake Barnby
ArnabChatterjee20k
loks0n