commit cefebc86af30522bf79d2a89a2bcf96f7f970ecb
author Alex Light <allight@google.com> Wed Jul 22 17:59:34 2020 -0700
committer Treehugger Robot <treehugger-gerrit@google.com> Thu Jul 23 17:08:49 2020 +0000
tree 6797788da75c2d6d0f4e26a883f25e2ba4e8a47a
parent c2208272f24e921b808af53721d2ca821d76503c

Print unicode data using java-encoding with -l xml

We would simply dump unicode bytes as-is in XML mode for dexdump. This
is often incorrect since xml is surprisingly picky about what chars
are valid. To avoid this whole headache simply encode all unicode
chars as you could for java strings. Also encode '\' as '\\' to avoid
ambiguity.

Test: manual
Test: dexdump -l xml out/target/product/blueline/testcases/CtsWebkitTestCases/arm64/CtsWebkitTestCases.apk | xmllint -
Bug: 161925303
Change-Id: Idc32d8257a98e173c6902c74c3788036f8683f16

dexdump/dexdump.cc

diff --git a/dexdump/dexdump.cc b/dexdump/dexdump.cc
index e694c5f..b6f3116 100644
--- a/dexdump/dexdump.cc
+++ b/dexdump/dexdump.cc
@@ -37,6 +37,8 @@
 #include <inttypes.h>
 #include <stdio.h>
 
+#include <cctype>
+#include <iomanip>
 #include <memory>
 #include <sstream>
 #include <vector>
@@ -45,6 +47,7 @@
 #include "android-base/logging.h"
 #include "android-base/stringprintf.h"
 
+#include "base/bit_utils.h"
 #include "dex/class_accessor-inl.h"
 #include "dex/code_item_accessors-inl.h"
 #include "dex/dex_file-inl.h"
@@ -360,10 +363,10 @@
 /*
  * Dumps a string value with some escape characters.
  */
-static void dumpEscapedString(const char* p) {
+static void dumpEscapedString(std::string_view p) {
   fputs("\"", gOutFile);
-  for (; *p; p++) {
-    switch (*p) {
+  for (char c : p) {
+    switch (c) {
       case '\\':
         fputs("\\\\", gOutFile);
         break;
@@ -380,42 +383,78 @@
         fputs("\\r", gOutFile);
         break;
       default:
-        putc(*p, gOutFile);
+        putc(c, gOutFile);
     }  // switch
   }  // for
   fputs("\"", gOutFile);
 }
 
+static size_t utf8Bytes(char start_byte) {
+  uint8_t sb = static_cast<uint8_t>(start_byte);
+  if ((sb & 0x80) == 0) {
+    return 1;
+  }
+  size_t msb = art::MostSignificantBit(static_cast<uint8_t>(~sb));
+  CHECK_LE(7u - msb, 4u);
+  return 7 - msb;
+}
+
 /*
  * Dumps a string as an XML attribute value.
  */
-static void dumpXmlAttribute(const char* p) {
-  for (; *p; p++) {
-    switch (*p) {
-      case '&':
-        fputs("&amp;", gOutFile);
-        break;
-      case '<':
-        fputs("&lt;", gOutFile);
-        break;
-      case '>':
-        fputs("&gt;", gOutFile);
-        break;
-      case '"':
-        fputs("&quot;", gOutFile);
-        break;
-      case '\t':
-        fputs("&#x9;", gOutFile);
-        break;
-      case '\n':
-        fputs("&#xA;", gOutFile);
-        break;
-      case '\r':
-        fputs("&#xD;", gOutFile);
-        break;
-      default:
-        putc(*p, gOutFile);
-    }  // switch
+static void dumpXmlAttribute(std::string_view p) __attribute__((optnone)) {
+  for (const char* c = p.begin(); c < p.end(); ++c) {
+    if (std::isprint(*c)) {
+      switch (*c) {
+        case '&':
+          fputs("&amp;", gOutFile);
+          break;
+        case '<':
+          fputs("&lt;", gOutFile);
+          break;
+        case '>':
+          fputs("&gt;", gOutFile);
+          break;
+        case '"':
+          fputs("&quot;", gOutFile);
+          break;
+        case '\\':
+          fputs("\\\\", gOutFile);
+          break;
+        default:
+          putc(*c, gOutFile);
+      }  // switch
+    } else {
+      uint32_t data = 0;
+      size_t remaining;
+      uint8_t uc = static_cast<uint8_t>(*c);
+      if (((uc) & 0x80) == 0) {
+        // Not a multi-byte char
+        data = static_cast<uint32_t>(*c);
+        remaining = 0;
+      } else if (utf8Bytes(uc) == 2) {
+        // 2 bytes
+        data = ((uc) & 0b00011111);
+        remaining = 1;
+      } else if (utf8Bytes(uc) == 3) {
+        // 3 bytes
+        data = ((uc) & 0b00001111);
+        remaining = 2;
+      } else {
+        // 4 bytes
+        CHECK_EQ(utf8Bytes(uc), 4u);
+        data = ((uc) & 0b00000111);
+        remaining = 3;
+      }
+      for (size_t i = 0; i < remaining; ++i) {
+        ++c;
+        data = data << 6;
+        uc = static_cast<uint8_t>(*c);
+        data |= static_cast<uint32_t>(uc & 0b00111111u);
+      }
+      // No good option so just use java encoding, too many chars are invalid
+      fprintf(gOutFile, "\\u%04x", data);
+    }
   }  // for
 }
 
@@ -478,9 +517,9 @@
     case DexFile::kDexAnnotationString: {
       const u4 idx = static_cast<u4>(readVarWidth(data, arg, false));
       if (gOptions.outputFormat == OUTPUT_PLAIN) {
-        dumpEscapedString(pDexFile->StringDataByIdx(dex::StringIndex(idx)));
+        dumpEscapedString(pDexFile->StringViewByIdx(dex::StringIndex(idx)));
       } else {
-        dumpXmlAttribute(pDexFile->StringDataByIdx(dex::StringIndex(idx)));
+        dumpXmlAttribute(pDexFile->StringViewByIdx(dex::StringIndex(idx)));
       }
       break;
     }
@@ -1666,7 +1705,7 @@
     fprintf(gOutFile, " target_class=\"%s\"\n", declaring_class);
     fprintf(gOutFile, " target_member=\"%s\"\n", member);
     fprintf(gOutFile, " target_member_type=");
-    dumpEscapedString(member_type.c_str());
+    dumpEscapedString(member_type);
     fprintf(gOutFile, "\n>\n</method_handle>\n");
   }
 }
@@ -1786,7 +1825,7 @@
       fprintf(gOutFile, "  link_argument[%zu] : %s (%s)\n", argument, value.c_str(), type);
     } else {
       fprintf(gOutFile, "<link_argument index=\"%zu\" type=\"%s\" value=", argument, type);
-      dumpEscapedString(value.c_str());
+      dumpEscapedString(value);
       fprintf(gOutFile, "/>\n");
     }