fix(data): reject stale reschedule snapshots

This commit is contained in:
2026-06-02 07:57:30 +03:00
parent 85918c1e5d
commit 1a8161027c
5 changed files with 97 additions and 5 deletions
@@ -80,6 +80,7 @@
- `d762ecc` `fix(data): serialize portfolio future reschedules`
- `1d62f69` `fix(data): lock racing portfolio publications`
- `ea71448` `fix(data): serialize new-link publication races`
- Current fix cycle: `fix(data): reject stale reschedule snapshots`
**Files:**
- Create: `tests/GmRelay.Bot.Tests/Web/PortfolioMigrationTests.cs`
@@ -271,6 +272,9 @@ public async Task ConcurrentPublishAndFutureReschedule_ShouldNotDeadlockOrCommit
[Fact]
public async Task ConcurrentNewLinkPublishAndFutureReschedule_ShouldNotCommitInvalidPublicCard()
[Fact]
public async Task RepeatableReadStaleSnapshotFutureReschedule_ShouldBeRejectedWithoutInvalidPublicCard()
[Theory]
[InlineData("portfolio_game_sessions")]
[InlineData("portfolio_game_masters")]
@@ -285,7 +289,7 @@ public async Task RequiredParentCascadeDelete_ShouldFailCommitForPublishedCard(s
public async Task ParentCardAndGroupCascadeDeletes_ShouldCommit()
```
The direct-delete, moved-link, invalid publication, and direct parent-cascade scenarios must expect PostgreSQL `23514` at commit. Every selected linked session must be completed with `scheduled_at < now()`: one future link among multiple selected sessions rejects publication. A final future reschedule must atomically unpublish linked public cards while preserving their first `published_at`; `past -> future -> past` in one transaction must leave the card public. Opposing-order batch reschedules must use an advisory test gate plus `pg_blocking_pids` observation with bounded timeouts, complete without card deadlock, and leave both cards private; do not rely on `pg_sleep` timing. The `READ COMMITTED` concurrency scenarios must launch bounded tasks together, cover both publish/delete lock orders, and prove there is no deadlock, write-skew, or invalid public commit. A session-delete versus future-reschedule race must use the common `sessions` then `portfolio_games` lock order, cover both first-session-lock orders through real blocking transactions, and finish with the card private and session deleted. The publish/reschedule races must finish with the future session committed and the card private, including a new-link draft publication forced behind the post-row-lock advisory gate. The `REPEATABLE READ` scenarios must reject triggered portfolio writes with `0A000`, including both draft-link deletion versus publication commit orders, because a stale snapshot after lock acquisition cannot safely validate the invariant. The parent-card and owning-group cascade scenarios must commit successfully.
The direct-delete, moved-link, invalid publication, and direct parent-cascade scenarios must expect PostgreSQL `23514` at commit. Every selected linked session must be completed with `scheduled_at < now()`: one future link among multiple selected sessions rejects publication. A final future reschedule must atomically unpublish linked public cards while preserving their first `published_at`; `past -> future -> past` in one transaction must leave the card public. Opposing-order batch reschedules must use an advisory test gate plus `pg_blocking_pids` observation with bounded timeouts, complete without card deadlock, and leave both cards private; do not rely on `pg_sleep` timing. The `READ COMMITTED` concurrency scenarios must launch bounded tasks together, cover both publish/delete lock orders, and prove there is no deadlock, write-skew, or invalid public commit. A session-delete versus future-reschedule race must use the common `sessions` then `portfolio_games` lock order, cover both first-session-lock orders through real blocking transactions, and finish with the card private and session deleted. The publish/reschedule races must finish with the future session committed and the card private, including a new-link draft publication forced behind the post-row-lock advisory gate. The `REPEATABLE READ` scenarios must reject triggered portfolio writes with `0A000`, including both draft-link deletion versus publication commit orders and stale-snapshot final-future reschedules after a newly linked publication, because a stale snapshot after lock acquisition cannot safely validate the invariant. The parent-card and owning-group cascade scenarios must commit successfully.
- [ ] **Step 3: Run the Task 1 tests to verify RED**
@@ -427,6 +431,12 @@ BEGIN
WHERE s.id = NEW.id;
IF final_scheduled_at >= now() THEN
IF current_setting('transaction_isolation') <> 'read committed' THEN
RAISE EXCEPTION
'portfolio future reschedule requires read committed isolation'
USING ERRCODE = '0A000';
END IF;
PERFORM pg.id
FROM portfolio_games pg
WHERE EXISTS (
@@ -514,7 +524,7 @@ CREATE INDEX ix_portfolio_game_reviews_pending
WHERE moderation_status = 'Pending';
```
The deferred constraint triggers retain the link-table `ON DELETE CASCADE` behavior. At transaction commit validators acquire the same transaction-level advisory lock and reject a surviving published card when either required link set is empty or any linked session has `scheduled_at >= now()`. The intentionally global lock is appropriate for low-volume portfolio publication writes: under the application default `READ COMMITTED` isolation level it serializes validation, prevents write-skew across distinct child links, and gives multi-card transactions one lock order. PostgreSQL retains stale snapshots under `REPEATABLE READ` and `SERIALIZABLE`, so the guard rejects every triggered portfolio write at those isolation levels with `0A000`. The deferred future-reschedule trigger re-reads the final session row, skips intermediate future values that end in the past, and for a final future value locks all cards linked to any final-future session in `portfolio_games.id` order. It then acquires the publication advisory lock and runs one guarded public-card unpublish update with a fresh `READ COMMITTED` statement snapshot. The row-lock phase deliberately includes committed drafts so a concurrent draft-to-public publication cannot pass validation against the pre-reschedule session snapshot and commit afterward; the post-row-lock advisory phase also serializes a previously invisible concurrent link-add publication without moving the advisory lock above card locks. At `READ COMMITTED`, draft edits, explicit unpublishing, future reschedules, and card or club cascade deletion remain valid. Normal session-deletion handlers use the same `sessions` then `portfolio_games` lock order: explicitly lock the target session row, unpublish linked cards, then delete the session.
The deferred constraint triggers retain the link-table `ON DELETE CASCADE` behavior. At transaction commit validators acquire the same transaction-level advisory lock and reject a surviving published card when either required link set is empty or any linked session has `scheduled_at >= now()`. The intentionally global lock is appropriate for low-volume portfolio publication writes: under the application default `READ COMMITTED` isolation level it serializes validation, prevents write-skew across distinct child links, and gives multi-card transactions one lock order. PostgreSQL retains stale snapshots under `REPEATABLE READ` and `SERIALIZABLE`, so the guard rejects every triggered portfolio write at those isolation levels with `0A000`. The deferred future-reschedule trigger re-reads the final session row, skips intermediate future values that end in the past, and rejects final-future reschedules outside `READ COMMITTED` with `0A000`. Under `READ COMMITTED`, it locks all cards linked to any final-future session in `portfolio_games.id` order. It then acquires the publication advisory lock and runs one guarded public-card unpublish update with a fresh statement snapshot. The row-lock phase deliberately includes committed drafts so a concurrent draft-to-public publication cannot pass validation against the pre-reschedule session snapshot and commit afterward; the post-row-lock advisory phase also serializes a previously invisible concurrent link-add publication without moving the advisory lock above card locks. At `READ COMMITTED`, draft edits, explicit unpublishing, future reschedules, and card or club cascade deletion remain valid. Normal session-deletion handlers use the same `sessions` then `portfolio_games` lock order: explicitly lock the target session row, unpublish linked cards, then delete the session.
- [ ] **Step 5: Lock sessions before explicitly unpublishing linked cards in both session-deletion handlers**
@@ -560,7 +570,7 @@ Run:
dotnet test tests/GmRelay.Bot.Tests/GmRelay.Bot.Tests.csproj --filter "FullyQualifiedName~PortfolioMigrationTests|FullyQualifiedName~PortfolioMigrationPostgresTests|FullyQualifiedName~PortfolioSessionDeletionSourceTests|FullyQualifiedName~PortfolioSchemaGateSourceTests"
```
Expected: PASS, including PostgreSQL 17 migration application, rejected direct required-link deletes, rejected moved links and session/player cascades, rejected publication with any future linked session, automatic unpublish with preserved `published_at` after final future reschedule, preserved public state after `past -> future -> past`, opposing-order batch reschedules without card deadlock, bounded `READ COMMITTED` publish/delete in both commit orders, existing-link and new-link publish/reschedule races, session-delete/reschedule serialization in both first-lock orders, and distinct-link deletion without deadlock, write-skew, or invalid public commit, rejected `REPEATABLE READ` triggered writes including both draft-delete versus publish commit orders, successful parent-card and owning-group cascades, Discord identity scoping, and Compose/Aspire HTTP health gating with a non-proxied bot endpoint and matching `gmrelaydb` resource name.
Expected: PASS, including PostgreSQL 17 migration application, rejected direct required-link deletes, rejected moved links and session/player cascades, rejected publication with any future linked session, automatic unpublish with preserved `published_at` after final future reschedule, preserved public state after `past -> future -> past`, opposing-order batch reschedules without card deadlock, bounded `READ COMMITTED` publish/delete in both commit orders, existing-link and new-link publish/reschedule races, session-delete/reschedule serialization in both first-lock orders, and distinct-link deletion without deadlock, write-skew, or invalid public commit, rejected `REPEATABLE READ` triggered writes including both draft-delete versus publish commit orders and stale-snapshot final-future reschedules, successful parent-card and owning-group cascades, Discord identity scoping, and Compose/Aspire HTTP health gating with a non-proxied bot endpoint and matching `gmrelaydb` resource name.
- [ ] **Step 7: Commit**
@@ -81,7 +81,7 @@ Application validation additionally requires at least one linked session, every
Deferred database constraint triggers validate the same invariant at transaction commit after a card transitions to public, a session link is inserted, deleted, moved, or repointed, or a required master link is deleted or moved. They raise a check-violation error if a published card would commit without both required link sets or with any linked session where `scheduled_at >= now()`. Before checking state, each validator acquires the same transaction-level PostgreSQL advisory lock, `pg_advisory_xact_lock(20260530, 108)`. Portfolio publication writes are low volume, so this intentionally global lock serializes invariant validation with one lock order, prevents write-skew under the application default `READ COMMITTED` isolation level, and avoids multi-card deadlocks. PostgreSQL keeps a stale snapshot after waiting under `REPEATABLE READ` or `SERIALIZABLE`, so the guard rejects every triggered portfolio write at those levels; callers must use `READ COMMITTED` for portfolio mutations.
A deferred `sessions.scheduled_at` trigger atomically unpublishes linked public cards when a completed session is finally rescheduled into the future, preserving the first `published_at`. Because deferred row triggers retain their event-time `NEW`, the trigger re-reads the final `sessions.scheduled_at` before acting. For a final future value it takes row locks for all cards linked to any final-future session in `portfolio_games.id` order, including committed drafts. It then acquires the publication advisory lock and unpublishes matching public cards in a guarded update with a fresh `READ COMMITTED` statement snapshot. Including drafts prevents a concurrent draft-to-public publication from validating against the pre-reschedule session snapshot and committing afterward. Taking the shared advisory lock after card rows, but before the guarded update, also serializes a previously invisible concurrent link-add publication without reintroducing the card/advisory lock inversion. Session mutation paths use `sessions` before linked `portfolio_games`; normal session-deletion handlers explicitly lock the target session row, unpublish linked cards in the same transaction, and only then delete the session. The link foreign keys retain `ON DELETE CASCADE`; when the card itself or its owning club is deleted at `READ COMMITTED`, deferred validation sees no surviving published card and remains harmless.
A deferred `sessions.scheduled_at` trigger atomically unpublishes linked public cards when a completed session is finally rescheduled into the future, preserving the first `published_at`. Because deferred row triggers retain their event-time `NEW`, the trigger re-reads the final `sessions.scheduled_at` before acting. It rejects final-future reschedules outside `READ COMMITTED` with `0A000`, because the unpublish pass requires fresh statement snapshots. Under `READ COMMITTED`, it takes row locks for all cards linked to any final-future session in `portfolio_games.id` order, including committed drafts. It then acquires the publication advisory lock and unpublishes matching public cards in a guarded update with a fresh statement snapshot. Including drafts prevents a concurrent draft-to-public publication from validating against the pre-reschedule session snapshot and committing afterward. Taking the shared advisory lock after card rows, but before the guarded update, also serializes a previously invisible concurrent link-add publication without reintroducing the card/advisory lock inversion. Session mutation paths use `sessions` before linked `portfolio_games`; normal session-deletion handlers explicitly lock the target session row, unpublish linked cards in the same transaction, and only then delete the session. The link foreign keys retain `ON DELETE CASCADE`; when the card itself or its owning club is deleted at `READ COMMITTED`, deferred validation sees no surviving published card and remains harmless.
### `portfolio_game_sessions`
@@ -356,7 +356,7 @@ Follow TDD for production changes.
### Schema And Contracts
- Migration source-contract tests assert the four new tables, format constraint, publication guard, case-insensitive slug uniqueness, group and GM-profile indexes, card-oriented pending-review index, completed-session validator, deferred future-reschedule unpublish trigger, session-first deletion locks, and the AppHost HTTP health gate.
- PostgreSQL integration tests apply migrations V001 through V029 to `postgres:17-alpine` and cover direct invalid link removal, moved links, direct session/player cascades, explicit session-lock then unpublish then session deletion, delete/reschedule lock ordering in both first-lock orders, rejection of publication when any linked session is future, automatic unpublish with preserved `published_at` after future reschedule, `past -> future -> past` final-state handling, opposing-order batch future reschedules without card deadlock using an observed advisory test gate instead of timing sleeps, existing-link and new-link draft publication/reschedule races, both bounded publish/delete commit orders, concurrent removal of distinct required links without write-skew or deadlock under `READ COMMITTED`, rejection of equivalent `REPEATABLE READ` writes including both draft-delete versus publish commit orders, and parent/card cascade deletion.
- PostgreSQL integration tests apply migrations V001 through V029 to `postgres:17-alpine` and cover direct invalid link removal, moved links, direct session/player cascades, explicit session-lock then unpublish then session deletion, delete/reschedule lock ordering in both first-lock orders, rejection of publication when any linked session is future, automatic unpublish with preserved `published_at` after future reschedule, `past -> future -> past` final-state handling, opposing-order batch future reschedules without card deadlock using an observed advisory test gate instead of timing sleeps, existing-link and new-link draft publication/reschedule races, both bounded publish/delete commit orders, concurrent removal of distinct required links without write-skew or deadlock under `READ COMMITTED`, rejection of equivalent `REPEATABLE READ` writes including both draft-delete versus publish commit orders and stale-snapshot final-future reschedules, and parent/card cascade deletion.
- Public DTO reflection/source tests assert that private identifiers and physical storage paths are absent.
- Existing showcase tests continue to assert the future-session catalog boundary.
@@ -128,6 +128,12 @@ BEGIN
WHERE s.id = NEW.id;
IF final_scheduled_at >= now() THEN
IF current_setting('transaction_isolation') <> 'read committed' THEN
RAISE EXCEPTION
'portfolio future reschedule requires read committed isolation'
USING ERRCODE = '0A000';
END IF;
PERFORM pg.id
FROM portfolio_games pg
WHERE EXISTS (
@@ -631,6 +631,81 @@ public sealed class PortfolioMigrationPostgresTests(PortfolioMigrationPostgresFi
]));
}
[Fact]
public async Task RepeatableReadStaleSnapshotFutureReschedule_ShouldBeRejectedWithoutInvalidPublicCard()
{
var database = await fixture.CreateMigratedDatabaseAsync();
await using var seedConnection = await database.OpenConnectionAsync();
var seed = await SeedCardAsync(seedConnection, isPublic: false);
var rescheduledSessionId = Guid.NewGuid();
await ExecuteNonQueryAsync(
seedConnection,
"""
INSERT INTO sessions (id, group_id, title, join_link, scheduled_at)
VALUES (@sessionId, @groupId, 'Completed Session', 'https://example.test/session', now() - interval '1 day');
""",
parameters:
[
new NpgsqlParameter("sessionId", rescheduledSessionId),
new NpgsqlParameter("groupId", seed.GroupId)
]);
await using var rescheduleConnection = await database.OpenConnectionAsync();
await using var publishConnection = await database.OpenConnectionAsync();
await using var rescheduleTransaction = await rescheduleConnection.BeginTransactionAsync(IsolationLevel.RepeatableRead);
Assert.Equal(0, await ExecuteScalarAsync<long>(
rescheduleConnection,
"""
SELECT COUNT(*)
FROM portfolio_game_sessions
WHERE portfolio_game_id = @portfolioGameId
AND session_id = @sessionId
""",
rescheduleTransaction,
new NpgsqlParameter("portfolioGameId", seed.PortfolioGameId),
new NpgsqlParameter("sessionId", rescheduledSessionId)));
await using (var publishTransaction = await publishConnection.BeginTransactionAsync())
{
await ExecuteNonQueryAsync(
publishConnection,
"""
INSERT INTO portfolio_game_sessions (portfolio_game_id, session_id)
VALUES (@portfolioGameId, @sessionId);
UPDATE portfolio_games
SET is_public = true,
published_at = COALESCE(published_at, now()),
updated_at = now()
WHERE id = @portfolioGameId
""",
publishTransaction,
new NpgsqlParameter("portfolioGameId", seed.PortfolioGameId),
new NpgsqlParameter("sessionId", rescheduledSessionId));
await publishTransaction.CommitAsync().WaitAsync(CommandTimeout);
}
Assert.Equal(1, await RescheduleSessionAsync(
rescheduleConnection,
rescheduleTransaction,
rescheduledSessionId));
var exception = await Assert.ThrowsAsync<PostgresException>(
() => rescheduleTransaction.CommitAsync().WaitAsync(CommandTimeout));
Assert.Equal(PostgresErrorCodes.FeatureNotSupported, exception.SqlState);
await using var verificationConnection = await database.OpenConnectionAsync();
Assert.True(await ExecuteScalarAsync<bool>(
verificationConnection,
"SELECT is_public FROM portfolio_games WHERE id = @portfolioGameId",
parameters: new NpgsqlParameter("portfolioGameId", seed.PortfolioGameId)));
Assert.False(await ExecuteScalarAsync<bool>(
verificationConnection,
"SELECT scheduled_at >= now() FROM sessions WHERE id = @sessionId",
parameters: new NpgsqlParameter("sessionId", rescheduledSessionId)));
}
[Theory]
[InlineData(true)]
[InlineData(false)]
@@ -48,6 +48,7 @@ public sealed class PortfolioMigrationTests
Assert.Contains("CREATE FUNCTION unpublish_public_portfolio_games_for_future_session() RETURNS TRIGGER LANGUAGE plpgsql", normalizedMigration, StringComparison.Ordinal);
Assert.Contains("SELECT s.scheduled_at INTO final_scheduled_at FROM sessions s WHERE s.id = NEW.id;", normalizedMigration, StringComparison.Ordinal);
Assert.Contains("IF final_scheduled_at >= now() THEN", normalizedMigration, StringComparison.Ordinal);
Assert.Contains("IF final_scheduled_at >= now() THEN IF current_setting('transaction_isolation') <> 'read committed' THEN RAISE EXCEPTION 'portfolio future reschedule requires read committed isolation' USING ERRCODE = '0A000'; END IF; PERFORM pg.id", normalizedMigration, StringComparison.Ordinal);
Assert.Contains("PERFORM pg.id FROM portfolio_games pg WHERE EXISTS", normalizedMigration, StringComparison.Ordinal);
Assert.Contains("ORDER BY pg.id FOR UPDATE OF pg;", normalizedMigration, StringComparison.Ordinal);
Assert.Contains("ORDER BY pg.id FOR UPDATE OF pg; PERFORM pg_advisory_xact_lock(20260530, 108); UPDATE portfolio_games pg SET is_public = false", normalizedMigration, StringComparison.Ordinal);