CONNE_IMPORT_WRONG_COMP_TYPE — IMPORT Structure Mismatch Runtime Error

IMPORT/EXPORT in ABAP does not move typed data directly; it serializes the source into a self-describing stream and later deserializes it into the target, checking that the component types line up field by field. CONNE_IMPORT_WRONG_COMP_TYPE fires during that deserialization step when a component in the stream has a type, length, or structure that the IMPORT target does not expect. It belongs to the CONNE (communication/connection) error class because the runtime treats the mismatch the same way it treats a broken RFC or file transfer: the data received does not match the contract the receiving side assumed. The dump is not a database error and not a user-data error in the business sense — it is a structural incompatibility between what was written and what is being read back, almost always introduced by a program or dictionary change that happened between the EXPORT and the IMPORT.

If the cause is a changed structure, the fix is either to convert the old cluster data to the new layout with a one-time conversion report, or to version the structure so old and new records can both be read (add fields only at the end, never insert or reorder) — this requires a change request and, for productive cluster tables, a data migration step. If the cause is memory-ID reuse, rename the memory ID to something unique per feature and re-test both producer and consumer together; this is a straightforward code change but still needs regression testing of every place that reads the old ID. If the cause is a transport-level inconsistency between systems or between a batch job and a later online step, the fix is procedural: align transport schedules so structure-changing transports and the jobs that depend on them move together, or add an explicit version tag to the exported payload so the importer can detect and reject stale data cleanly instead of crashing. If the cause is an SAP-delivered structure changed by an upgrade, check whether SAP ships a conversion program for the cluster table before writing a custom one.

The reflex fix is adding the IGNORING STRUCTURE BOUNDARIES addition to the IMPORT statement, or wrapping the IMPORT in a TRY/CATCH that swallows the exception and moves on. Both suppress the crash but not the mismatch: IGNORING STRUCTURE BOUNDARIES tells the runtime to map bytes across field boundaries regardless of type, which silently populates fields with garbage or shifted values instead of failing loudly. A caught exception that is logged and ignored leaves the program running on an incomplete or wrongly typed data object, which produces incorrect business results discovered much later and far harder to trace than the original dump.

Never reuse a literal memory ID string across unrelated features; prefix it with the program or function group name. Treat any structure used in EXPORT/IMPORT to a persistent cluster table as a versioned interface: only append fields at the end, document every consumer, and write a conversion report before any structural change reaches production. For interfaces between systems on different release levels, exchange an explicit version indicator in the payload rather than relying on implicit layout agreement. Include IMPORT/EXPORT usage in code review checklists whenever a structure feeding a cluster table is touched.

This is an ABAP development issue, not a Basis or functional one, except when it surfaces after a transport sequencing problem across systems, in which case Basis needs to confirm which transports landed where and when. The handover note should list the memory ID or cluster table involved, the exporting and importing program names and line numbers, the structure definition before and after the suspected change, and the transport request that introduced the change.

Seen alongside CALL_FUNCTION_CONFLICT_TYPE and CALL_FUNCTION_CONFLICT_LENG when the same kind of structural drift affects RFC interface parameters instead of EXPORT/IMPORT statements, and alongside MOVE_CAST_ERROR when the mismatch is discovered via a dynamic cast rather than a data cluster read. Involves database cluster tables (INDX-style, including custom Z cluster tables), ABAP memory (SET/GET PARAMETER is a different mechanism and not affected), and any program pair that shares data through EXPORT/IMPORT rather than a typed interface.