fix(ucs): use NO_ERROR_CODE fallback when connector error code is absent

[AmitsinghTanwar007]

Summary

When the UCS psync path returns a connector HTTP error (4xx/5xx) with no connector-specific error code, decode_connector_error_response was falling back to the gRPC discriminator string "CONNECTOR_ERROR_RESPONSE" instead of "No error code". This caused a shadow-validation diff on response.Err.code between the HS direct path and the UCS path for connectors (e.g. cybersource psync) that return error responses without a specific error code.

Diff signature: router.valueDiff:response.Err.code
Layer: HS→UCS mapping (hyperswitch_interfaces/src/unified_connector_service/transformers.rs)
Flow: cybersource / psync

Root cause

When a connector returns a 4xx error with no specific error code:

1. HS direct path → build_error_response → code = "No error code" (the NO_ERROR_CODE constant)
2. UCS path → same build_error_response → code = "No error code" → ForeignFrom<&ErrorResponse>: since code == NO_ERROR_CODE AND reason == None, has_connector_details = false → error_info = None in the gRPC ConnectorError message → HS decode_connector_error_response: connector_details is None → the .unwrap_or_else(|| connector_error.error_code.clone()) fallback returned "CONNECTOR_ERROR_RESPONSE" ← wrong

The fix changes the fallback from connector_error.error_code.clone() to crate::consts::NO_ERROR_CODE.to_string() so both paths surface "No error code" when the connector provides no specific code.

Verification

Verified by source-parity analysis (cybersource TLS-blocked locally):

Before fix: decode_connector_error_response fallback → code = "CONNECTOR_ERROR_RESPONSE" → HS shadow diff: response.Err.code = "No error code" (HS) vs "CONNECTOR_ERROR_RESPONSE" (UCS)

After fix: decode_connector_error_response fallback → code = "No error code" → both paths agree → no_diff

Unit tests added:

• decode_prefers_connector_details_code: when connector_details.code = Some("INVALID_REQUEST") → surfaced code is "INVALID_REQUEST"
• decode_falls_back_to_no_error_code_when_connector_code_absent: when error_info = None → surfaced code is "No error code"

Closes #13094

[semanticdiff-com]

Review changes with  

[XyneSpaces]

⚠️ Error code fallback logic may mask connector-specific failures

The change from connector_error.error_code.clone() to crate::consts::NO_ERROR_CODE.to_string() as a fallback when connector_details is absent may mask legitimate connector errors in logging/monitoring.

Consider preserving the original error context even when falling back:

let error_code = connector_details
    .as_ref()
    .map(|d| d.error_code.clone())
    .unwrap_or_else(|| {
        logger::warn!("Connector error code unavailable, using NO_ERROR_CODE fallback");
        crate::consts::NO_ERROR_CODE.to_string()
    });

This ensures operational visibility into when connectors return HTTP errors without specific error codes.

[XyneSpaces]

📝 Code Review: UCS Error Code Fallback

This is a straightforward fix ensuring consistent error code behavior between UCS and direct HS paths.

✅ Change Analysis

The modification replaces a fallback that would return "CONNECTOR_ERROR_RESPONSE" with the standardized "No error code" constant. This improves consistency across the codebase.

💡 Minor Suggestion

Consider adding a brief comment explaining why NO_ERROR_CODE is preferred over the connector's raw error response, for future maintainers.

🟢 LGTM

Simple, correct fix with no blocking issues.

[XyneSpaces]

💡 The NO_ERROR_CODE and NO_ERROR_MESSAGE constants used as fallbacks should have corresponding metrics/alerts when triggered. Consider adding a counter metric to track how frequently the fallback is used, as this indicates incomplete error-code mapping.

Also verify that downstream systems (Control Center, analytics) handle these sentinel values gracefully.

[XyneSpaces]

[nit] This PR is a clean, targeted fix. The change from connector_error.error_code.clone() to NO_ERROR_CODE correctly prevents leaking the internal gRPC discriminator "CONNECTOR_ERROR_RESPONSE" to end users. Tests cover both branches appropriately. No changes requested.

[XyneSpaces]

💡 Test coverage is good

The unit tests added for the NO_ERROR_CODE fallback are well-structured and cover both the happy path (connector provides code) and fallback path. Good practice adding these tests alongside the fix.