Use the original MIT troubleshooting kit with the free text proto checker. No account is required for this checker. Its synthetic tables demonstrate diagnosis; they are not complete engine-ready protos.
Start from the known-good fixture
Extract the kit and select both files from fixed/: item_proto.txt and item_names.txt. Run the check. Expected result: zero errors and zero warnings. Keep these originals unchanged and make a separate working copy for each exercise.
Exercise A: a single ID inside a range
The fixture includes the range 2000~2003. Append a new row to a copy of item_proto.txt, using actual tab separators:
2002 Overlap example 0Select the modified proto and the unchanged names file. Expected result: one error, zero warnings. The checker reports VNUM 2002 overlapping the range on line 4; the added row is line 5. Changing the name does not remove the numeric overlap. Remove the added row and rerun: expect zero findings again.
Exercise B: a missing target is a different problem
Start again with the original fixed files. In the row for VNUM 1000, change only REFINED_VNUM from 1001 to 9999. Expected result: one error, zero warnings. Line 2 reports that refine target 9999 is absent. Restore 1001 and rerun: expect zero findings. Do not add an arbitrary production item just to silence a diagnostic; determine the intended upgrade chain first.
What these results prove
Both mutations were run against the current M2City checker on 3 November 2026. The expected counts and line numbers above matched. This verifies these synthetic parser cases, not quest compilation, binary proto decoding or game compatibility. The refine check depends on the recognised column header; headerless layouts are not guessed.
Try a different case and discuss it
Can you reduce your own confusing report to a similarly small synthetic example? Reply with filename, encoding, header, only the necessary rows, expected result and actual JSON report. State whether you started from exercise A or B. Preserve tabs when sharing data and remove private names. A report that identifies one unsupported layout is more useful than an unexplained full server pack.
The fixture and checker are original M2City resources under MIT. You can share the exercise with attribution and answer follow-up questions in the community where you share it.