| Previous | Next |
| CO_E_WRONGOSFORAPP | CO_E_OBJISREG |
CO_E_OBJNOTREG
Meaning and numerical form
CO_E_OBJNOTREG is HRESULT 2147746299 (0x800401FB; signed -2147220997). AllStat records its meaning as “Object is not registered”. In practical terms, it is returned when an object-registration operation expects a live object entry but no matching registration is present.
The severity bit in 0x800401FB is set, so this result is a failure result rather than a success or informational status. Its facility is 4 (FACILITY_ITF) and its low 16-bit code is 507 (0x01FB). Those fields classify the value but do not identify the component instance, call, object or input that produced it.
Evidence to collect
A useful this result record includes the native API or interface method, component or CLSID, process and thread IDs, architecture, apartment where relevant, operation correlation ID, elapsed time, input identity, and the first state-changing event before failure. When diagnosing it, preserve IErrorInfo, an inner HRESULT, Win32 status, device event or provider message when available.
- Capture 1 for it: the exact registration API, identity key and registration token used by the producer.
- Capture 2 for it: timestamps for registration, revocation and the failed lookup.
- Capture 3 for it: process lifetime and running-object-table state in the same user session.
Telemetry for it should retain types, lengths, hashes, stable opaque identifiers and lifecycle generations while redacting content and credentials. This makes repeated incidents comparable without exposing paths, object identities, user data or credentials.
Where the failure belongs
This result belongs to COM activation and object lifetime. COM activation is a sequence rather than one registry lookup: identifier resolution, class registration, module or process start, class-factory publication, marshaling, and object lifetime can fail independently. The decisive condition is this: an object-registration operation expects a live object entry but no matching registration is present. Record the first native producer instead of relying only on a framework exception translated several layers later.
Do not infer from this result alone that reinstalling, rebooting or repeating the whole workflow is appropriate. Because it can cross process and language boundaries, its correction depends on the object identity, lifecycle generation, input and state at the first failing call.
Likely causes
- Possible cause 1 for it: the producer never registered the object instance.
- Possible cause 2 for it: the registration was revoked during shutdown or recovery.
- Possible cause 3 for it: the lookup key, moniker or object identity differs from the value used at registration.
Diagnostic sequence
- Capture it at the first native return and identify the exact object and method.
- Confirm that the observed path matches this the condition: an object-registration operation expects a live object entry but no matching registration is present.
- compare the live state with the method contract before changing configuration.
- collect the code-specific evidence listed above and preserve any more specific preceding error.
- change one evidence-backed cause at a time, then repeat the smallest operation that can prove the correction.
- Add a regression check that distinguishes it from the related outcomes described below.
Correction and retry boundary
- Corrective action 1 for it: register the object before exposing discovery to clients.
- Corrective action 2 for it: use the same canonical identity for registration and lookup.
- Corrective action 3 for it: serialize shutdown so revocation cannot race an in-flight consumer.
Retry guidance for it. Retry after a producer has created a new registration; blind polling should be bounded and tied to observable server startup. A it retry must avoid this hazard: launch duplicate servers, reuse stale proxies or repeat an activation whose side effects are unknown. It must also preserve the evidence needed to locate the original producer of it.
Difference from nearby HRESULTs
REGDB_E_CLASSNOTREG concerns class activation metadata, while it concerns a missing live object registration. Keep these outcomes separate in logs, exception mappings, user messages, dashboards and automated retry policy.
Developer and operational guidance
Code handling it should log the failed stage before cleanup, keep ownership and lifecycle transitions explicit, and avoid converting every failure into a generic message. Tests for it should cover the primary cause, one adjacent HRESULT and the retry boundary so a future change cannot turn a deterministic contract failure into an uncontrolled loop.
Operational repair for it should use supported product installation, COM registration, activation and class-factory lifetime paths. Avoid unsupported configuration edits or component replacement unless evidence identifies that layer. Retain the original trace and a post-fix trace so the result can be attributed and the change reversed.
Practical scenario
A document server revokes its running-object entry during restart while a client still looks up the old moniker; waiting for the new registration resolves it. The result scenario narrows the result to a reproducible contract boundary and supplies a concrete before-and-after test.
Official Microsoft references
Looking for a different code? Search another status or error code.