Ghidra script improvements (#7)

* Always set image base to 0 for Ghidra

* Script improvements

Everything here is for Ghidra:
* Use `%` instead of f-string (Ghidra still uses python 2.7, meanwhile f-strings were added in 3.6)
* Handle errors when applying type (Ghidra throws exception unlike IDA)
* Don't trigger decompiler (analysis will be faster)
* Revert back string literals

* Set image base to zero only for ELF

I don't know about PE with Ghidra

* Update README for disassemblers

* IDA 7.6 required due `ida_ida.inf_is_32bit_exactly()`

* Ghidra now don't launch decompiler for whole binary

* Set image base to 0 in script

Remove %IMAGE_BASE% since we don't use it anymore

* Create XREFs for Ghidra

Now you can jump from `MethodInfo` to actual method

* Fix demangler for Ghidra

Ghidra's demangler can process only functions in auto analysis. Now both `TypeInfo` and `MethodInfo` are displayed properly

---------

Co-authored-by: commonuserlol <commonuserlol@users.noreply.github.com>

Il2CppInspector.Common/Outputs/ScriptResources/Targets/Ghidra.py

@@ -4,9 +4,19 @@ from ghidra.app.script import GhidraScriptUtil
 from ghidra.app.util.cparser.C import CParserUtils
 from ghidra.program.model.data import ArrayDataType
 from ghidra.program.model.symbol import SourceType
+from ghidra.program.model.symbol import RefType
+from ghidra.app.cmd.label import DemanglerCmd
+
+xrefs = currentProgram.getReferenceManager()
 
 def set_name(addr, name):
-	createLabel(toAddr(addr), name, True)
+	if not name.startswith("_ZN"):
+		createLabel(toAddr(addr), name, True)
+		return
+	cmd = DemanglerCmd(currentAddress.getAddress(hex(addr)), name)
+	if not cmd.applyTo(currentProgram, monitor):
+		print("Failed to apply demangled name to %s at %s due %s, falling back to mangled" % (name, hex(addr), cmd.getStatusMsg()))
+		createLabel(toAddr(addr), name, True)
 
 def make_function(start, end = None):
 	addr = toAddr(start)
@@ -41,10 +51,13 @@ def set_type(addr, cppType):
 	if cppType.startswith('struct '):
 		cppType = cppType[7:]
 	
-	t = getDataTypes(cppType)[0]
-	addr = toAddr(addr)
-	removeDataAt(addr)
-	createData(addr, t)
+	try:
+		t = getDataTypes(cppType)[0]
+		addr = toAddr(addr)
+		removeDataAt(addr)
+		createData(addr, t)
+	except:
+		print("Failed to set type: %s" % cppType)
 
 def set_comment(addr, text):
 	setEOLComment(toAddr(addr), text)
@@ -61,15 +74,33 @@ def script_prologue(status):
 
 	# Ghidra sets the image base for ELF to 0x100000 for some reason
 	# https://github.com/NationalSecurityAgency/ghidra/issues/1020
+	# Make sure that the base address is 0
+	# Without this, Ghidra may not analyze the binary correctly and you will just waste your time
+	# If 0 doesn't work for you, replace it with the base address from the output of the CLI or GUI
 	if currentProgram.getExecutableFormat().endswith('(ELF)'):
-		currentProgram.setImageBase(toAddr(%IMAGE_BASE%), True)
+		currentProgram.setImageBase(toAddr(0), True)
+	
+	# Don't trigger decompiler
+	setAnalysisOption(currentProgram, "Call Convention ID", "false")
 
 def get_script_directory(): return getSourceFile().getParentFile().toString()
 
 def script_epilogue(status): pass
 def add_function_to_group(addr, group): pass
-def add_xref(addr, to): pass
-def create_fake_segment(name, size): pass
-def write_string(addr, string): pass
-def write_address(addr, value): pass
+def add_xref(addr, to):
+	xrefs.addMemoryReference(currentAddress.getAddress(hex(addr)), currentAddress.getAddress(hex(to)), RefType.DATA, SourceType.USER_DEFINED, 0)
+
+def process_string_literals(status, data):
+	for d in jsonData['stringLiterals']:
+		define_string(d)
+
+		# I don't know how to make inline strings in Ghidra
+		# Just revert back original impl
+		addr = parse_address(d)
+		set_name(addr, d['name'])
+		set_type(addr, r'struct String *')
+		set_comment(addr, d['string'])
+
+		status.update_progress()
+
 class StatusHandler(BaseStatusHandler): pass