From bc0df62e20f77f411bc971c2c7b2e0994a271b06 Mon Sep 17 00:00:00 2001
From: jonschz <jonschz@users.noreply.github.com>
Date: Thu, 11 Jul 2024 19:34:31 +0200
Subject: [PATCH] Spike to fix array comparisons (needs refactor)

---
 tools/isledecomp/isledecomp/compare/core.py   | 70 ++++++++++++++++++-
 .../isledecomp/isledecomp/cvdump/analysis.py  |  5 +-
 2 files changed, 73 insertions(+), 2 deletions(-)

diff --git a/tools/isledecomp/isledecomp/compare/core.py b/tools/isledecomp/isledecomp/compare/core.py
index 1587ef81..d7f99d58 100644
--- a/tools/isledecomp/isledecomp/compare/core.py
+++ b/tools/isledecomp/isledecomp/compare/core.py
@@ -8,6 +8,7 @@
 from isledecomp.bin import Bin as IsleBin, InvalidVirtualAddressError
 from isledecomp.cvdump.demangler import demangle_string_const
 from isledecomp.cvdump import Cvdump, CvdumpAnalysis
+from isledecomp.cvdump.types import scalar_type_pointer
 from isledecomp.parser import DecompCodebase
 from isledecomp.dir import walk_source_dir
 from isledecomp.types import SymbolType
@@ -220,7 +221,62 @@ def orig_bin_checker(addr: int) -> bool:
                     var.offset, var.name, var.parent_function
                 )
             else:
-                self._db.match_variable(var.offset, var.name)
+                if not self._db.match_variable(var.offset, var.name):
+                    continue
+                # retrieve compare address
+                matchinfo = self._db.get_by_orig(var.offset)
+                if matchinfo is None or matchinfo.recomp_addr is None:
+                    continue
+
+                var_recomp_addr = matchinfo.recomp_addr
+
+                node = next(
+                    (
+                        x
+                        for x in self.cvdump_analysis.nodes
+                        if x.addr == var_recomp_addr
+                    ),
+                    None,
+                )
+                if node is None or node.data_type is None:
+                    continue
+
+                if not node.data_type.key.startswith("0x"):
+                    # scalar type, no further processing needed
+                    continue
+
+                data_type = self.cv.types.keys[node.data_type.key.lower()]
+
+                if data_type["type"] == "LF_ARRAY":
+                    array_size_bytes = data_type["size"]
+
+                    array_type = self.cv.types.get(data_type["array_type"])
+
+                    assert array_type.size is not None
+                    array_type_size = array_type.size
+
+                    array_length, modulus = divmod(array_size_bytes, array_type_size)
+                    assert modulus == 0
+
+                    # 0 is already matched when we get here
+                    for i in range(1, array_length):
+                        orig_element_base_addr = var.offset + i * array_type_size
+                        recomp_element_base_addr = var_recomp_addr + i * array_type_size
+                        if array_type.members is None:
+                            self._add_match_in_array(
+                                f"{var.name}[{i}]",
+                                array_type.key,
+                                orig_element_base_addr,
+                                recomp_element_base_addr,
+                            )
+                        else:
+                            for member in array_type.members:
+                                self._add_match_in_array(
+                                    f"{var.name}[{i}].{member.name}",
+                                    array_type.key,
+                                    orig_element_base_addr + member.offset,
+                                    recomp_element_base_addr + member.offset,
+                                )
 
         for tbl in codebase.iter_vtables():
             self._db.match_vtable(tbl.offset, tbl.name, tbl.base_class)
@@ -245,6 +301,18 @@ def orig_bin_checker(addr: int) -> bool:
 
             self._db.match_string(string.offset, string.name)
 
+    def _add_match_in_array(
+        self, name: str, type_id: str, orig_addr: int, recomp_addr: int
+    ):
+        self._db.set_recomp_symbol(
+            recomp_addr,
+            SymbolType.POINTER if scalar_type_pointer(type_id) else SymbolType.DATA,
+            name,
+            name,
+            4,
+        )
+        self._db.set_pair(orig_addr, recomp_addr)
+
     def _find_original_strings(self):
         """Go to the original binary and look for the specified string constants
         to find a match. This is a (relatively) expensive operation so we only
diff --git a/tools/isledecomp/isledecomp/cvdump/analysis.py b/tools/isledecomp/isledecomp/cvdump/analysis.py
index 40ef292e..e030035e 100644
--- a/tools/isledecomp/isledecomp/cvdump/analysis.py
+++ b/tools/isledecomp/isledecomp/cvdump/analysis.py
@@ -5,7 +5,7 @@
 from isledecomp.types import SymbolType
 from .parser import CvdumpParser
 from .demangler import demangle_string_const, demangle_vtable
-from .types import CvdumpKeyError, CvdumpIntegrityError
+from .types import CvdumpKeyError, CvdumpIntegrityError, TypeInfo
 
 
 class CvdumpNode:
@@ -35,6 +35,8 @@ class CvdumpNode:
     section_contribution: Optional[int] = None
     addr: Optional[int] = None
     symbol_entry: Optional[SymbolsEntry] = None
+    # Preliminary - only used for non-static variables at the moment
+    data_type: Optional[TypeInfo] = None
 
     def __init__(self, section: int, offset: int) -> None:
         self.section = section
@@ -127,6 +129,7 @@ def __init__(self, parser: CvdumpParser):
                 # get information for built-in "T_" types.
                 g_info = parser.types.get(glo.type)
                 node_dict[key].confirmed_size = g_info.size
+                node_dict[key].data_type = g_info
                 # Previously we set the symbol type to POINTER here if
                 # the variable was known to be a pointer. We can derive this
                 # information later when it's time to compare the variable,