New Zip::ZipFromMemory API.

third_party/zlib/google/zip.cc

Index: third_party/zlib/google/zip.cc
diff --git a/third_party/zlib/google/zip.cc b/third_party/zlib/google/zip.cc
index 52fc6948bb6aa38a68c5dec74284786d3faca707..a6661272938b41e75578b8eecd16e3075c2b5178 100644
--- a/third_party/zlib/google/zip.cc
+++ b/third_party/zlib/google/zip.cc
@@ -11,6 +11,7 @@
 #include "base/file_util.h"
 #include "base/files/file_enumerator.h"
 #include "base/logging.h"
+#include "base/memory/ref_counted_memory.h"
 #include "base/strings/string16.h"
 #include "base/strings/string_util.h"
 #include "net/base/file_stream.h"
@@ -84,30 +85,32 @@ bool AddFileToZip(zipFile zip_file, const base::FilePath& src_dir) {
   return true;
 }
 
-bool AddEntryToZip(zipFile zip_file, const base::FilePath& path,
-                   const base::FilePath& root_path) {
-  base::FilePath relative_path;
-  bool result = root_path.AppendRelativePath(path, &relative_path);
-  DCHECK(result);
-  std::string str_path = relative_path.AsUTF8Unsafe();
-#if defined(OS_WIN)
-  ReplaceSubstringsAfterOffset(&str_path, 0u, "\\", "/");
-#endif
+bool AddMemContentsToZip(zipFile zip_file,
+                         const base::FilePath& file_path,
+                         base::RefCountedMemory* contents) {
+  DCHECK(contents);
+  if (contents->size() != 0 &&
+      ZIP_OK != zipWriteInFileInZip(
+          zip_file, contents->front(), contents->size())) {
+    DLOG(ERROR) << "Could not write data to zip for path "
+                << file_path.value();
+    return false;
+  }
 
-  bool is_directory = base::DirectoryExists(path);
-  if (is_directory)
-    str_path += "/";
+  return true;
+}
 
+bool ZipOpenNewFileWrapper(zipFile zip_file,
+                           const std::string& str_path,
+                           const zip_fileinfo* file_info) {
   // Section 4.4.4 http://www.pkware.com/documents/casestudies/APPNOTE.TXT
   // Setting the Language encoding flag so the file is told to be in utf-8.
   const uLong LANGUAGE_ENCODING_FLAG = 0x1 << 11;
 
-  zip_fileinfo file_info = GetFileInfoForZipping(path);
-
   if (ZIP_OK != zipOpenNewFileInZip4(
                     zip_file,  // file
                     str_path.c_str(),  // filename
-                    &file_info,  // zipfi
+                    file_info,  // zipfi
                     NULL,  // extrafield_local,
                     0u,  // size_extrafield_local
                     NULL,  // extrafield_global
@@ -126,6 +129,26 @@ bool AddEntryToZip(zipFile zip_file, const base::FilePath& path,
     DLOG(ERROR) << "Could not open zip file entry " << str_path;
     return false;
   }
+  return true;
+}
+
+bool AddEntryToZip(zipFile zip_file, const base::FilePath& path,
+                   const base::FilePath& root_path) {
+  base::FilePath relative_path;
+  bool result = root_path.AppendRelativePath(path, &relative_path);
+  DCHECK(result);
+  std::string str_path = relative_path.AsUTF8Unsafe();
+#if defined(OS_WIN)
+  ReplaceSubstringsAfterOffset(&str_path, 0u, "\\", "/");
+#endif
+
+  bool is_directory = base::DirectoryExists(path);
+  if (is_directory)
+    str_path += "/";
+
+  zip_fileinfo file_info = GetFileInfoForZipping(path);
+  if (!ZipOpenNewFileWrapper(zip_file, str_path, &file_info))
+    return false;
 
   bool success = true;
   if (!is_directory) {
@@ -148,6 +171,133 @@ bool ExcludeHiddenFilesFilter(const base::FilePath& file_path) {
   return file_path.BaseName().value()[0] != '.';
 }
 
+// Tells if there is at least one file in the |zip_path| archive which also
+// exists in |contents|.
+// Returns false if there was a problem opening the zip file.
+// The result of whether there is an intersection between the paths in
+// |zip_path| and |contents| is stored in |has_files|.
+bool HasFileInZip(const base::FilePath& zip_path,
+                  const zip::ZipContents& contents,
+                  bool* has_files) {
+  DCHECK(has_files);
+  *has_files = false;
+
+  zip::ZipReader reader;
+  if (!reader.Open(zip_path)) {
+    DLOG(ERROR) << "Can't open zip file '" << zip_path.value();
+    return false;
+  }
+
+  for (zip::ZipContents::const_iterator itr = contents.begin();
+       itr != contents.end(); ++itr)
+    if (reader.LocateAndOpenEntry(itr->first)) {
+      *has_files = true;
+      return true;
+    }
+
+  return true;
+}
+
+// Validates the contents in a zip::ZipContents to be used by ZipFromMemory:
+// Makes sure that all paths are relative and safe. Verifies that folders do not
+// have data (such would perhaps be a bug when utilizing the API).
+bool ValidateZipContents(const zip::ZipContents& contents) {
+  for (zip::ZipContents::const_iterator itr = contents.begin();
+       itr != contents.end(); ++itr) {
+    if (itr->first.empty() ||
+        itr->first.IsAbsolute() ||
+        itr->first.ReferencesParent()) {
+      DLOG(ERROR) << "Invalid file path '" << itr->first.value() << "'";
+      return false;
+    }
+    if (itr->first.EndsWithSeparator() &&
+        itr->second.get() &&
+        itr->second.get()->size() != 0) {
+      // Can be NULL (remove) or 0-byte (add).
+      DLOG(ERROR) << "Folder must have empty contents '" <<
+          itr->first.value() << "'";
+      return false;
+    }
+  }
+  return true;
+}
+
+// Creates a new file or folder entry in |zip_file| with |path| as the file
+// path and |contents| as the file contents. If |path| ends with a separator
+// it is considered to be a folder entry. Returns false in case of error.
+bool AddEntryToZipFromMemory(zipFile zip_file,
+                             const base::FilePath& path,
+                             base::RefCountedMemory* contents) {
+  DCHECK(!path.IsAbsolute());
+  std::string str_path = path.AsUTF8Unsafe();
+#if defined(OS_WIN)
+  ReplaceSubstringsAfterOffset(&str_path, 0u, "\\", "/");
+#endif
+
+  zip_fileinfo file_info = TimeToZipFileInfo(base::Time::Now());
+  if (!ZipOpenNewFileWrapper(zip_file, str_path, &file_info))
+    return false;
+
+  bool success = true;
+  if (!path.EndsWithSeparator()) {
+    // Not a folder.
+    DCHECK(contents);
+    success = AddMemContentsToZip(zip_file, path, contents);
+  }
+
+  if (ZIP_OK != zipCloseFileInZip(zip_file)) {
+    DLOG(ERROR) << "Could not close zip file entry " << str_path;
+    return false;
+  }
+
+  return success;
+}
+
+// Finds all files (path and contents) present in |zip_path| which are not
+// present in |contents| and adds them to |output_zip_file|
+// The function returns false in case of error handling the zip file.
+bool CopyZipContentsNotInMap(const base::FilePath& zip_path,
+                             const zip::ZipContents& contents,
+                             zipFile output_zip_file) {
+  // Arbitrary big size, but not that big, of the maximum amount of memory this
+  // function can use to load a file's contents into memory.
+  const int64 max_allowed_file_size = 128 * 1024 * 1024;
+
+  zip::ZipContents other_contents;
+
+  zip::ZipReader reader;
+  if (!reader.Open(zip_path)) {
+    DLOG(ERROR) << "Can't open zip file '" << zip_path.value();
+    return false;
+  }
+
+  while (reader.HasMore()) {
+    if (!reader.OpenCurrentEntryInZip()) {
+      DLOG(ERROR) << "Can't open entry in zip file '" << zip_path.value();
+      return false;
+    }
+
+    zip::ZipReader::EntryInfo* entry = reader.current_entry_info();
+    DCHECK(entry);
+    if (contents.find(entry->file_path()) == contents.end()) {
+      scoped_refptr<base::RefCountedMemory> mem;
+      if (!reader.ExtractCurrentEntryToRefCountedMemory(
+               max_allowed_file_size, &mem)) {
+        DLOG(ERROR) << "Failed to read contents of "
+                    << entry->file_path().value();
+        return false;
+      }
+      if (!AddEntryToZipFromMemory(
+               output_zip_file, entry->file_path(), mem.get()))
+        return false;
+    }
+    reader.AdvanceToNextEntry();
+  }
+  reader.Close();
+
+  return true;
+}
+
 }  // namespace
 
 namespace zip {
@@ -197,7 +347,8 @@ bool ZipWithFilterCallback(const base::FilePath& src_dir,
   bool success = true;
   base::FileEnumerator file_enumerator(src_dir, true /* recursive */,
       base::FileEnumerator::FILES | base::FileEnumerator::DIRECTORIES);
-  for (base::FilePath path = file_enumerator.Next(); !path.value().empty();
+  for (base::FilePath path = file_enumerator.Next();
+      !path.value().empty() && success;
        path = file_enumerator.Next()) {
     if (!filter_cb.Run(path)) {
       continue;
@@ -205,7 +356,6 @@ bool ZipWithFilterCallback(const base::FilePath& src_dir,
 
     if (!AddEntryToZip(zip_file, path, src_dir)) {
       success = false;
-      return false;
     }
   }
 
@@ -243,7 +393,7 @@ bool ZipFiles(const base::FilePath& src_dir,
   bool success = true;
   for (std::vector<base::FilePath>::const_iterator iter =
            src_relative_paths.begin();
-      iter != src_relative_paths.end(); ++iter) {
+      iter != src_relative_paths.end() && success; ++iter) {
     const base::FilePath& path = src_dir.Append(*iter);
     if (!AddEntryToZip(zip_file, path, src_dir)) {
       // TODO(hshi): clean up the partial zip file when error occurs.
@@ -261,4 +411,94 @@ bool ZipFiles(const base::FilePath& src_dir,
 }
 #endif  // defined(OS_POSIX)
 
+bool ZipFromMemory(const base::FilePath& zip_path,
+                   const ZipContents& contents,
+                   bool append) {
+  DCHECK(!zip_path.empty());
+
+  // There is really no point in calling this with no input, and
+  // there are simpler ways to create a 0 byte file.
+  if (contents.size() == 0)
+    return true;
+
+  // This check saves us having to worry with invalid input after.
+  if (!ValidateZipContents(contents))
+    return false;
+
+  int64 file_size = 0;
+  if (append &&
+      (!base::PathExists(zip_path) ||
+       (base::GetFileSize(zip_path, &file_size) && file_size < 1))) {
+    // If the size is 0, this might be a new temporary file.
+    append = false;
+  }
+
+  // The path to the zip path that will actually be used to write the new data.
+  base::FilePath zip_path_used = zip_path;
+
+  bool has_intersection = false;
+  if (append) {
+    // NOTE: zipOpenNewFileInZip4 is not clever enough to replace files,
+    // therefore this has to explicitly create a new file and reinsert the
+    // data we want without duplications, which is what the minizip headers
+    // tell to do.
+    if (!HasFileInZip(zip_path, contents, &has_intersection))
+      return false;
+
+    if (has_intersection) {
+      append = false;
+      if (!base::CreateTemporaryFile(&zip_path_used)) {
+        DLOG(WARNING) << "Couldn't create temporary file for "
+                      << zip_path.value();
+        return false;
+      }
+    }
+  }
+
+  zipFile zip_file = internal::OpenForZipping(
+      zip_path_used.AsUTF8Unsafe(),
+      append ? APPEND_STATUS_ADDINZIP : APPEND_STATUS_CREATE);
+  if (!zip_file) {
+    DLOG(WARNING) << "Couldn't create file " << zip_path_used.value();
+    return false;
+  }
+
+  bool success = true;
+
+  if (has_intersection) {
+    // There was an intersection. There are files in the old archive with
+    // same paths as found in |contents| therefore the old ones need to be
+    // replaced.
+    success = CopyZipContentsNotInMap(zip_path, contents, zip_file);
+  }
+
+  for (ZipContents::const_iterator itr = contents.begin();
+       itr != contents.end() && success;
+       ++itr) {
+    if (itr->second.get()) {
+      success = AddEntryToZipFromMemory(
+          zip_file, itr->first, itr->second.get());
+    }
+  }
+
+  if (ZIP_OK != zipClose(zip_file, NULL)) {
+    DLOG(ERROR) << "Error closing zip file " << zip_path_used.value();
+    success = false;
+  }
+
+  if (zip_path_used != zip_path) {
+    if (success) {
+      if (!base::Move(zip_path_used, zip_path)) {
+        base::DeleteFile(zip_path_used, false);
+        DLOG(ERROR) << "Error moving zip file " << zip_path_used.value();
+        return false;
+      }
+    } else {
+      base::DeleteFile(zip_path_used, false);
+    }
+  }
+
+  return success;
+}
+
 }  // namespace zip