Merge fix/secrets-prehost-content-root: run the store-path guard where the store is actually created
ci / nightly-windev (push) Has been skipped
ci / windows-x86 (push) Successful in 1m17s
ci / java (push) Successful in 3m12s
ci / portable (push) Successful in 10m36s

0.6.0 landed the rules but not the enforcement on this path: the pre-host secrets
container has no IHostEnvironment, so the library skipped the content-root check
and the migrator created the database before the real host could refuse it. The
pin to 0.6.1 alone does not close that — the call site has to pass the content
root explicitly, which is why this is a code change and not a version bump.
This commit is contained in:
Joseph Doherty
2026-08-12 02:13:40 -04:00
3 changed files with 125 additions and 4 deletions
@@ -78,9 +78,19 @@ public static class GatewayApplication
// here (SecretNotFoundException); config with no tokens is untouched (no-op), so this is safe // here (SecretNotFoundException); config with no tokens is untouched (no-op), so this is safe
// to always run. CreateBuilder is synchronous and single-shot at bootstrap, so the two awaits // to always run. CreateBuilder is synchronous and single-shot at bootstrap, so the two awaits
// are driven via GetAwaiter().GetResult() (no sync-context deadlock risk during host startup). // are driven via GetAwaiter().GetResult() (no sync-context deadlock risk during host startup).
// The content root is passed explicitly because this container is a throwaway
// ServiceCollection with no IHostEnvironment in it. Without it the library cannot tell
// "no content root exists" from "no host is registered", so it skips the
// under-content-root rule — and the migrator below CREATES the store before the real host
// ever validates. The boot then fails a moment later, having already left an empty
// database with its -wal/-shm siblings at the very path the rule rejects. That artifact is
// what made the 2026-08-09 outage read as "the database is there, it's just empty".
// DO NOT simplify this to the 3-argument overload: it still compiles, the app still boots
// when the path is correct, and the guard silently stops running at the one moment that
// matters.
#pragma warning disable ASP0000 // deliberate throwaway container, disposed here, shares no singletons #pragma warning disable ASP0000 // deliberate throwaway container, disposed here, shares no singletons
using (var secretsProvider = new ServiceCollection() using (var secretsProvider = new ServiceCollection()
.AddZbSecrets(builder.Configuration, "Secrets") .AddZbSecrets(builder.Configuration, "Secrets", builder.Environment.ContentRootPath)
.BuildServiceProvider()) .BuildServiceProvider())
#pragma warning restore ASP0000 #pragma warning restore ASP0000
{ {
@@ -21,9 +21,9 @@
<PackageReference Include="ZB.MOM.WW.Telemetry" Version="0.1.0" /> <PackageReference Include="ZB.MOM.WW.Telemetry" Version="0.1.0" />
<PackageReference Include="ZB.MOM.WW.Telemetry.Serilog" Version="0.1.0" /> <PackageReference Include="ZB.MOM.WW.Telemetry.Serilog" Version="0.1.0" />
<PackageReference Include="ZB.MOM.WW.GalaxyRepository" Version="0.2.0" /> <PackageReference Include="ZB.MOM.WW.GalaxyRepository" Version="0.2.0" />
<PackageReference Include="ZB.MOM.WW.Secrets" Version="0.6.0" /> <PackageReference Include="ZB.MOM.WW.Secrets" Version="0.6.1" />
<PackageReference Include="ZB.MOM.WW.Secrets.Abstractions" Version="0.6.0" /> <PackageReference Include="ZB.MOM.WW.Secrets.Abstractions" Version="0.6.1" />
<PackageReference Include="ZB.MOM.WW.Secrets.Ui" Version="0.6.0" /> <PackageReference Include="ZB.MOM.WW.Secrets.Ui" Version="0.6.1" />
<PackageReference Include="Serilog.AspNetCore" Version="10.0.0" /> <PackageReference Include="Serilog.AspNetCore" Version="10.0.0" />
<PackageReference Include="Serilog.Sinks.Console" Version="6.1.1" /> <PackageReference Include="Serilog.Sinks.Console" Version="6.1.1" />
<PackageReference Include="Serilog.Sinks.File" Version="7.0.0" /> <PackageReference Include="Serilog.Sinks.File" Version="7.0.0" />
@@ -0,0 +1,111 @@
using Microsoft.AspNetCore.Builder;
using ZB.MOM.WW.MxGateway.Server;
namespace ZB.MOM.WW.MxGateway.Tests.Configuration;
// Mutates the process-global Secrets__SqlitePath that GatewayApplication.CreateBuilder reads;
// serialized against every other collection so a parallel host-building test cannot inherit the
// deliberately-rejected path. See GlobalEnvironmentCollection.
[Collection(TestSupport.GlobalEnvironmentCollection.Name)]
public sealed class SecretsStorePathGuardTests
{
private const string SqlitePathVariable = "Secrets__SqlitePath";
/// <summary>
/// Verifies the store-path guard runs in the <em>pre-host</em> secrets container, which is the
/// only place it matters.
/// </summary>
/// <remarks>
/// <para>
/// <c>CreateBuilder</c> resolves <c>${secret:}</c> references before the host exists, using a
/// throwaway <see cref="Microsoft.Extensions.DependencyInjection.ServiceCollection"/> that
/// contains no <c>IHostEnvironment</c> — and it runs the store migrator, which <b>creates the
/// database</b>. A library that infers the content root from <c>IHostEnvironment</c> alone
/// cannot distinguish "no content root" from "no host registered" and skips the rule here, so
/// the store is created at the rejected path and only then does the real host refuse to start.
/// The leftover empty database with its <c>-wal</c>/<c>-shm</c> siblings is precisely the
/// artifact that made the 2026-08-09 credential loss read as "the database is there, it's just
/// empty". The gateway therefore passes the content root explicitly.
/// </para>
/// <para>
/// The assertion that no file was created is the load-bearing one. A test that merely observed
/// a failed boot would pass even while the store was being written, because the failure arrives
/// afterwards either way — which is exactly how this defect survived its first release.
/// </para>
/// </remarks>
[Fact]
public void CreateBuilder_RejectsSecretsStoreUnderContentRoot_WithoutCreatingIt()
{
string? original = Environment.GetEnvironmentVariable(SqlitePathVariable);
string contentRoot = ResolveContentRoot();
string rejected = Path.Combine(contentRoot, $"probe-secrets-{Guid.NewGuid():N}.db");
try
{
Environment.SetEnvironmentVariable(SqlitePathVariable, rejected);
// Capture rather than Assert.ThrowsAny, so the store-creation assertions below are
// reported first. Ordering matters here: "it threw" is the weaker claim, and asserting
// it first would mask the stronger one — that nothing was written before it threw.
Exception? thrown = Record.Exception(() => GatewayApplication.CreateBuilder([]));
Assert.False(File.Exists(rejected), $"the rejected store was created at {rejected}");
Assert.False(File.Exists(rejected + "-wal"), "a write-ahead log was created for the rejected store");
Assert.False(File.Exists(rejected + "-shm"), "a shared-memory file was created for the rejected store");
Assert.NotNull(thrown);
}
finally
{
Environment.SetEnvironmentVariable(SqlitePathVariable, original);
// Delete defensively: if the guard ever regresses this test writes a database into the
// content root, which on a dev machine is the source tree.
foreach (string leftover in new[] { rejected, rejected + "-wal", rejected + "-shm" })
{
if (File.Exists(leftover))
{
File.Delete(leftover);
}
}
}
}
/// <summary>
/// Verifies a store path outside the content root is accepted <em>and actually used</em>.
/// </summary>
/// <remarks>
/// The assertion is that the database exists afterwards, not merely that nothing threw. A
/// not-null builder is very close to a tautology once no exception escaped, so it would pass
/// even if the pre-host container had stopped opening the store altogether — which would also
/// silently void the negative test above, since that one can only observe a file the migration
/// would otherwise have written. Proving the accepted path gets a real database is what keeps
/// the rejected-path assertion meaningful.
/// </remarks>
[Fact]
public void CreateBuilder_AcceptsSecretsStoreOutsideContentRoot_AndCreatesIt()
{
string? original = Environment.GetEnvironmentVariable(SqlitePathVariable);
string directory = Directory.CreateTempSubdirectory("mxgw-secrets-ok").FullName;
string accepted = Path.Combine(directory, "secrets.db");
try
{
Environment.SetEnvironmentVariable(SqlitePathVariable, accepted);
WebApplicationBuilder builder = GatewayApplication.CreateBuilder([]);
Assert.NotNull(builder);
Assert.True(File.Exists(accepted), $"the accepted store was not created at {accepted}");
}
finally
{
Environment.SetEnvironmentVariable(SqlitePathVariable, original);
Directory.Delete(directory, recursive: true);
}
}
// The content root CreateBuilder will use, taken from a builder created with the suite's normal
// (valid) store path rather than assumed from the test's working directory.
private static string ResolveContentRoot() =>
GatewayApplication.CreateBuilder([]).Environment.ContentRootPath;
}