ghidra-rename-backup

name	ghidra-rename-backup
description	Backup and restore Ghidra function/label renames to/from a JSON file in the git repo
version	"2026-04-13T00:00:00.000Z"

Ghidra Rename Backup & Restore

CRITICAL: Always backup renames after a RE session. Restoring from backup is the ONLY way to recover if the Ghidra project DB gets corrupted or re-imported.

Architecture

• Renames are exported to ~/hamsterball-re/analysis/ghidra/renames.json
• Format: {"renames": [{"address": "00401234", "old_name": "FUN_00401234", "new_name": "Ball_Update"}], ...}
• This file is committed to git so renames survive project DB loss

Export Renames (run after each session batch)

# The export is done via GhidraMCP - run the inline script
# Or use the MCP batch_create_labels tool

Export via Inline Ghidra Script

Use mcp_ghidra_mcp_run_script_inline with this Java code:

import ghidra.program.model.symbol.*;
import ghidra.program.model.listing.*;
import java.util.*;

var sb = new StringBuilder();
sb.append("{\n  \"renames\": [\n");
var fm = currentProgram.getFunctionManager();
var iter = fm.getFunctions(true);
var first = true;
while (iter.hasNext()) {
    var func = iter.next();
    var name = func.getName();
    if (!name.startsWith("FUN_") && !name.startsWith("LAB_")) {
        if (!first) sb.append(",\n");
        first = false;
        sb.append("    {\"address\": \"")
          .append(Long.toHexString(func.getEntryPoint().getOffset()))
          .append("\", \"new_name\": \"")
          .append(name.replace("\"", "\\\""))
          .append("\"}");
    }
}
// Also get non-function labels
var symTable = currentProgram.getSymbolTable();
var symIter = symTable.getSymbolIterator();
while (symIter.hasNext()) {
    var sym = symIter.next();
    var name = sym.getName();
    var addr = sym.getAddress();
    if (!name.startsWith("FUN_") && !name.startsWith("LAB_") && !name.startsWith("DAT_") && !name.startsWith("EXT_") && !name.startsWith("Start") && !name.startsWith("thunk_") && fm.getFunctionContaining(addr) == null) {
        if (!first) sb.append(",\n");
        first = false;
        sb.append("    {\"address\": \"")
          .append(Long.toHexString(addr.getOffset()))
          .append("\", \"new_name\": \"")
          .append(name.replace("\"", "\\\""))
          .append("\"}");
    }
}
sb.append("\n  ]\n}");
println(sb.toString());

Capture the output and save to ~/hamsterball-re/analysis/ghidra/renames.json.

Restore Renames (run after Ghidra project re-import)

1. Read ~/hamsterball-re/analysis/ghidra/renames.json
2. Parse the JSON array of renames
3. For each entry, call mcp_ghidra_mcp_rename_or_label with the address and new_name
4. Process in batches of ~50 to avoid timeout issues

Batch Restore Script

import json

with open('/home/evan/hamsterball-re/analysis/ghidra/renames.json') as f:
    data = json.load(f)

renames = data['renames']
print(f"Total renames to apply: {len(renames)}")

# Split into batches of 50
batches = [renames[i:i+50] for i in range(0, len(renames), 50)]
print(f"Batches: {len(batches)}")

Then for each batch, call mcp_ghidra_mcp_batch_create_labels with the labels array.

Pitfalls

• The Ghidra project DB is NOT version-controlled. Always export renames to JSON.
• When using analyzeHeadless to re-import, all previous renames are LOST unless restored from backup.
• The headless MCP server needs --program /Hamsterball.exe (with leading slash) to load a project program.
• Export after every rename batch, or at minimum after each session.

Quick Reference

Action	Command
Export renames	Run inline script, save output to renames.json
Restore renames	Read renames.json, batch_create_labels
Verify count	mcp_ghidra_mcp_compare_programs_documentation
Commit to git	cd ~/hamsterball-re && git add analysis/ghidra/renames.json && git commit -m "backup renames"