Skip to main content

Troubleshooting

EAppInjector raised at startup​

Symptom: TNidus.Create raises EAppInjector.

Cause: GNidusInject is nil — the Nidus.Inject unit was not loaded before GetNidus was called.

Fix: Add Nidus.Inject to the uses clause of your project's .dpr file (or any unit that is loaded before the first call to GetNidus). The initialization section of Nidus.Inject creates GNidusInject.

// myproject.dpr
uses
  Nidus.Inject,   // ← must appear early
  Nidus,
  ...

EModuleStartedException raised​

Symptom: TNidus.Start raises EModuleStartedException.

Cause: Start was called more than once on the same TNidus instance.

Fix: Call Start exactly once per application lifetime. If you need to restart, call Finalize first and then obtain a fresh TNidus via GetNidus (after re-creating the inject container).

HTTP 401 on every request​

Symptom: All requests return HTTP 401 Unauthorized immediately after registering a guard.

Cause: The TGuardCallback always returns False, or it accesses GetNidus.Request before the request is assigned.

Fix:

.UseGuard(
  function: Boolean
  var LReq: IRouteRequest;
  begin
    LReq := GetNidus.Request;
    if LReq = nil then
      Exit(False);
    Result := ValidateToken(LReq.Authorization);
  end)

HTTP 400 with validation errors​

Symptom: Route returns HTTP 400 even though the payload looks valid.

Cause: The validation pipe found one or more failing decorator constraints on the DTO.

Fix:

  1. Check the JSON body key names match the DTO property names exactly — the lookup is case-sensitive (TJsonParserMap preserves key casing, and TValidationPipe builds the cache key as ClassName->PropertyName using the RTTI property name as-is).
  2. The msg parameter in every decorator attribute is optional (default = ''); both [IsEmail] and [IsEmail('custom message')] are valid. When msg is empty, the validator generates a default message from the tag name, type name, field name, and value.
  3. Temporarily disable UsePipes to isolate whether the error comes from validation.

ERPCProviderNotSetException raised​

Symptom: PublishRPC raises ERPCProviderNotSetException.

Cause: UseRPC(...) was not called before PublishRPC.

Fix:

GetNidus
  .UseRPC(TRPCServerIndy.Create('0.0.0.0', 9900))
  .PublishRPC('MyMethod', TMyRPCResource)
  .Start(TAppModule.Create);

Module instances not freed between requests​

Symptom: Memory grows across requests; module constructors only called once.

Cause: Module caching is enabled for the module (either via SetPolicy or EnableAll).

Fix: Remove the module from the cache policy or call GetNidus.Cache.Invalidate(TYourModule) to evict it.

Linux64: OutputDebugString linker error​

Symptom: Linker error mentioning OutputDebugString when building for Linux64.

Cause: An older version of Nidus.Exception called OutputDebugString unconditionally.

Fix: Update to the current source. The DEBUG trace in Nidus.Exception is now wrapped in {$IFDEF MSWINDOWS} with a stderr fallback on POSIX. No code change required in your application.

Nidus_Horse route not dispatching​

Symptom: Horse routes are not reached — all requests return empty or error.

Cause: The Nidus middleware was registered after the route handlers, so it never calls Next.

Fix: Ensure Nidus_Horse middleware is added before route handlers:

THorse
  .Use(Nidus_Horse(TAppModule.Create))  // ← middleware first
  .Get('/users', HandleGetUsers)
  .Listen(9000);

Pool Acquire always returns a new instance​

Symptom: TObjectPool<T>.Count stays at 0; every Acquire creates a new object.

Cause: Release is never called, so objects are never returned to the pool.

Fix: Use TNidus.WithPool<T> which guarantees release in all paths, or wrap manual acquire/release in try/finally:

GetNidus.WithPool<TMyResource>(
  procedure(R: TMyResource)
  begin
    R.DoWork;
  end);