Add get_string_data to DexIdx and use in DexLoader

Summary: Do not preload DexString instances when checking type ids by using the new method.

Reviewed By: thezhangwei

Differential Revision: D67782728

fbshipit-source-id: 2ad71d9114f6e21137ceb0ebe48a27ddc135fd9a

libredex/DexIdx.cpp

@@ -115,7 +115,8 @@ DexMethodHandle* DexIdx::get_methodhandleidx_fromdex(uint32_t mhidx) {
   }
 }
 
-const DexString* DexIdx::get_stringidx_fromdex(uint32_t stridx) {
+std::string_view DexIdx::get_string_data(uint32_t stridx,
+                                         uint32_t* utfsize) const {
   redex_assert(stridx < m_string_ids_size);
   uint32_t stroff = m_string_ids[stridx].offset;
   // Bounds check is conservative. May incorrectly reject short strings
@@ -125,17 +126,23 @@ const DexString* DexIdx::get_stringidx_fromdex(uint32_t stridx) {
   const uint8_t* dstr = m_dexbase + stroff;
   /* Strip off uleb128 size encoding */
 
-  uint32_t utfsize = read_uleb128(&dstr);
+  uint32_t utfsize_local = read_uleb128(&dstr);
+  if (utfsize != nullptr) {
+    *utfsize = utfsize_local;
+  }
   // Find null terminator.
   auto null_cur = dstr;
   while (*null_cur != '\0' && null_cur < m_dexbase + get_file_size()) {
     ++null_cur;
   }
   always_assert_type_log(null_cur < m_dexbase + get_file_size(), INVALID_DEX,
                          "Missing null terminator");
-
-  auto ret = DexString::make_string(
-      std::string_view((const char*)dstr, null_cur - dstr));
+  return std::string_view((const char*)dstr, null_cur - dstr);
+}
+const DexString* DexIdx::get_stringidx_fromdex(uint32_t stridx) {
+  uint32_t utfsize;
+  auto str_data = get_string_data(stridx, &utfsize);
+  auto ret = DexString::make_string(str_data);
   always_assert_type_log(
       ret->length() == utfsize,
       INVALID_DEX,

libredex/DexIdx.h

@@ -65,6 +65,7 @@ class DexIdx {
  public:
   explicit DexIdx(const dex_header* dh);
 
+  std::string_view get_string_data(uint32_t stridx, uint32_t* utfsize) const;
   const DexString* get_stringidx(uint32_t stridx) {
     always_assert_type_log(
         stridx < m_string_ids_size, RedexError::INVALID_DEX,

libredex/DexLoader.cpp

@@ -255,12 +255,11 @@ void validate_type_ids_table(DexIdx* idx, const dex_header* dh) {
     always_assert_type_log(type_id.string_idx < str_size, INVALID_DEX,
                            "Type index out of bounds");
 
-    // This will preload the string, but otherwise we duplicate some nontrivial
-    // decode checking.
-    auto* dex_str = idx->get_stringidx(type_id.string_idx);
-    always_assert_type_log(type::is_valid(dex_str->str()), INVALID_DEX,
-                           "%s is not a valid type descriptor",
-                           dex_str->c_str());
+    // Don't preload the string, just check the plain data.
+    auto dex_str =
+        idx->get_string_data(type_id.string_idx, /*utfsize=*/nullptr);
+    always_assert_type_log(type::is_valid(dex_str), INVALID_DEX,
+                           "%s is not a valid type descriptor", dex_str.data());
   }
 }