[net] Cleanup filename_util and make it portable.

net/base/filename_util_internal.cc

Index: net/base/filename_util_internal.cc
diff --git a/net/base/filename_util_internal.cc b/net/base/filename_util_internal.cc
index c77375b40d30afa2d75cda645e40ee37fffd799a..f26b83552461f6e79df7fabde1dc795f618e327d 100644
--- a/net/base/filename_util_internal.cc
+++ b/net/base/filename_util_internal.cc
@@ -14,76 +14,22 @@
 #include "net/base/filename_util_internal.h"
 #include "net/base/mime_util.h"
 #include "net/base/net_string_util.h"
+#include "net/base/net_util.h"
 #include "net/http/http_content_disposition.h"
 #include "url/gurl.h"
 
 namespace net {
 
-void SanitizeGeneratedFileName(base::FilePath::StringType* filename,
-                               bool replace_trailing) {
-  const base::FilePath::CharType kReplace[] = FILE_PATH_LITERAL("-");
-  if (filename->empty())
-    return;
-  if (replace_trailing) {
-    // Handle CreateFile() stripping trailing dots and spaces on filenames
-    // http://support.microsoft.com/kb/115827
-    size_t length = filename->size();
-    size_t pos = filename->find_last_not_of(FILE_PATH_LITERAL(" ."));
-    filename->resize((pos == std::string::npos) ? 0 : (pos + 1));
-    base::TrimWhitespace(*filename, base::TRIM_TRAILING, filename);
-    if (filename->empty())
-      return;
-    size_t trimmed = length - filename->size();
-    if (trimmed)
-      filename->insert(filename->end(), trimmed, kReplace[0]);
-  }
-  base::TrimString(*filename, FILE_PATH_LITERAL("."), filename);
-  if (filename->empty())
-    return;
-  // Replace any path information by changing path separators.
-  ReplaceSubstringsAfterOffset(filename, 0, FILE_PATH_LITERAL("/"), kReplace);
-  ReplaceSubstringsAfterOffset(filename, 0, FILE_PATH_LITERAL("\\"), kReplace);
-}
-
-// Returns the filename determined from the last component of the path portion
-// of the URL.  Returns an empty string if the URL doesn't have a path or is
-// invalid. If the generated filename is not reliable,
-// |should_overwrite_extension| will be set to true, in which case a better
-// extension should be determined based on the content type.
-std::string GetFileNameFromURL(const GURL& url,
-                               const std::string& referrer_charset,
-                               bool* should_overwrite_extension) {
-  // about: and data: URLs don't have file names, but esp. data: URLs may
-  // contain parts that look like ones (i.e., contain a slash).  Therefore we
-  // don't attempt to divine a file name out of them.
-  if (!url.is_valid() || url.SchemeIs("about") || url.SchemeIs("data"))
-    return std::string();
-
-  const std::string unescaped_url_filename = UnescapeURLComponent(
-      url.ExtractFileName(),
-      UnescapeRule::SPACES | UnescapeRule::URL_SPECIAL_CHARS);
+// Our replacement character is an underscore. It used to be a '-', but that's
+// problematic since if we replace a character at the beginning of a string, the
+// resulting filename could look like a command line option when used on the
+// command line on Posix systems.
+const char kReplacement = '_';
 
-  // The URL's path should be escaped UTF-8, but may not be.
-  std::string decoded_filename = unescaped_url_filename;
-  if (!base::IsStringUTF8(decoded_filename)) {
-    // TODO(jshin): this is probably not robust enough. To be sure, we need
-    // encoding detection.
-    base::string16 utf16_output;
-    if (!referrer_charset.empty() &&
-        net::ConvertToUTF16(
-            unescaped_url_filename, referrer_charset.c_str(), &utf16_output)) {
-      decoded_filename = base::UTF16ToUTF8(utf16_output);
-    } else {
-      decoded_filename =
-          base::WideToUTF8(base::SysNativeMBToWide(unescaped_url_filename));
-    }
-  }
-  // If the URL contains a (possibly empty) query, assume it is a generator, and
-  // allow the determined extension to be overwritten.
-  *should_overwrite_extension = !decoded_filename.empty() && url.has_query();
+namespace {
 
-  return decoded_filename;
-}
+const base::FilePath::CharType kDefaultExtension[] =
+    FILE_PATH_LITERAL("download");
 
 // Returns whether the specified extension is automatically integrated into the
 // windows shell.
@@ -106,7 +52,7 @@ bool IsShellIntegratedExtension(const base::FilePath::StringType& extension) {
   return false;
 }
 
-// Returns whether the specified file name is a reserved name on windows.
+// Returns whether the specified file name is a reserved name on Windows.
 // This includes names like "com2.zip" (which correspond to devices) and
 // desktop.ini and thumbs.db which have special meaning to the windows shell.
 bool IsReservedName(const base::FilePath::StringType& filename) {
@@ -114,7 +60,7 @@ bool IsReservedName(const base::FilePath::StringType& filename) {
   // http://msdn2.microsoft.com/en-us/library/aa365247(VS.85).aspx
   // I also added clock$ because GetSaveFileName seems to consider it as a
   // reserved name too.
-  static const char* const known_devices[] = {
+  static const char* const kKnownDeviceNames[] = {
       "con",  "prn",  "aux",  "nul",  "com1", "com2", "com3",  "com4",
       "com5", "com6", "com7", "com8", "com9", "lpt1", "lpt2",  "lpt3",
       "lpt4", "lpt5", "lpt6", "lpt7", "lpt8", "lpt9", "clock$"};
@@ -125,203 +71,240 @@ bool IsReservedName(const base::FilePath::StringType& filename) {
   std::string filename_lower = base::StringToLowerASCII(filename);
 #endif
 
-  for (size_t i = 0; i < arraysize(known_devices); ++i) {
+  for (const auto& device_name : kKnownDeviceNames) {
     // Exact match.
-    if (filename_lower == known_devices[i])
+    if (filename_lower == device_name)
       return true;
     // Starts with "DEVICE.".
-    if (filename_lower.find(std::string(known_devices[i]) + ".") == 0)
+    if (filename_lower.find(std::string(device_name) + ".") == 0)
       return true;
   }
 
-  static const char* const magic_names[] = {
-    // These file names are used by the "Customize folder" feature of the shell.
-    "desktop.ini",
-    "thumbs.db",
+  static const char* const kMagicNames[] = {
+      // These file names are used by the "Customize folder" feature of the
+      // shell.
+      "desktop.ini",
+      "thumbs.db",
   };
 
-  for (size_t i = 0; i < arraysize(magic_names); ++i) {
-    if (filename_lower == magic_names[i])
+  for (const auto& magic_name : kMagicNames) {
+    if (filename_lower == magic_name)
       return true;
   }
 
   return false;
 }
 
-// Examines the current extension in |file_name| and modifies it if necessary in
-// order to ensure the filename is safe.  If |file_name| doesn't contain an
-// extension or if |ignore_extension| is true, then a new extension will be
-// constructed based on the |mime_type|.
-//
-// We're addressing two things here:
-//
-// 1) Usability.  If there is no reliable file extension, we want to guess a
-//    reasonable file extension based on the content type.
-//
-// 2) Shell integration.  Some file extensions automatically integrate with the
-//    shell.  We block these extensions to prevent a malicious web site from
-//    integrating with the user's shell.
-void EnsureSafeExtension(const std::string& mime_type,
-                         bool ignore_extension,
-                         base::FilePath* file_name) {
-  // See if our file name already contains an extension.
-  base::FilePath::StringType extension = file_name->Extension();
-  if (!extension.empty())
-    extension.erase(extension.begin());  // Erase preceding '.'.
+// Returns the UTF-8 filename determined from the last component of the path
+// portion of the URL.  Returns an empty string if the URL doesn't have a path
+// or is invalid. Updates |extension_option| if a filename can be extracted from
+// the URL.
+std::string GetFileNameFromURL(const GURL& url,
+                               const std::string& referrer_charset,
+                               ExtensionGenerationOption* extension_option) {
+  // about: and data: URLs don't have file names, but esp. data: URLs may
+  // contain parts that look like ones (i.e., contain a slash).  Therefore we
+  // don't attempt to divine a file name out of them.
+  if (!url.is_valid() || url.SchemeIs("about") || url.SchemeIs("data"))
+    return std::string();
 
-  if ((ignore_extension || extension.empty()) && !mime_type.empty()) {
-    base::FilePath::StringType preferred_mime_extension;
-    std::vector<base::FilePath::StringType> all_mime_extensions;
-    net::GetPreferredExtensionForMimeType(mime_type, &preferred_mime_extension);
-    net::GetExtensionsForMimeType(mime_type, &all_mime_extensions);
-    // If the existing extension is in the list of valid extensions for the
-    // given type, use it. This avoids doing things like pointlessly renaming
-    // "foo.jpg" to "foo.jpeg".
-    if (std::find(all_mime_extensions.begin(),
-                  all_mime_extensions.end(),
-                  extension) != all_mime_extensions.end()) {
-      // leave |extension| alone
-    } else if (!preferred_mime_extension.empty()) {
-      extension = preferred_mime_extension;
+  const std::string unescaped_url_filename = UnescapeURLComponent(
+      url.ExtractFileName(),
+      UnescapeRule::SPACES | UnescapeRule::URL_SPECIAL_CHARS);
+
+  // The URL's path should be escaped UTF-8, but may not be.
+  std::string decoded_filename = unescaped_url_filename;
+  if (!base::IsStringUTF8(decoded_filename)) {
+    // TODO(jshin): this is probably not robust enough. To be sure, we need
+    // encoding detection.
+    base::string16 utf16_output;
+    if (!referrer_charset.empty() &&
+        net::ConvertToUTF16(unescaped_url_filename, referrer_charset.c_str(),
+                            &utf16_output)) {
+      decoded_filename = base::UTF16ToUTF8(utf16_output);
+    } else {
+      decoded_filename =
+          base::WideToUTF8(base::SysNativeMBToWide(unescaped_url_filename));
     }
   }
+  // If the URL contains a (possibly empty) query, assume it is a generator, and
+  // allow the determined extension to be overwritten.
+  if (!decoded_filename.empty()) {
+    *extension_option =
+        (url.has_query() ? EXTENSION_OPTION_GENERATE
+                         : EXTENSION_OPTION_GENERATE_IF_MISSING);
+  }
 
-#if defined(OS_WIN)
-  static const base::FilePath::CharType default_extension[] =
-      FILE_PATH_LITERAL("download");
-
-  // Rename shell-integrated extensions.
-  // TODO(asanka): Consider stripping out the bad extension and replacing it
-  // with the preferred extension for the MIME type if one is available.
-  if (IsShellIntegratedExtension(extension))
-    extension.assign(default_extension);
-#endif
-
-  *file_name = file_name->ReplaceExtension(extension);
+  return decoded_filename;
 }
 
-bool FilePathToString16(const base::FilePath& path, base::string16* converted) {
-#if defined(OS_WIN)
-  *converted = path.value();
-  return true;
-#elif defined(OS_POSIX)
-  std::string component8 = path.AsUTF8Unsafe();
-  return !component8.empty() &&
-         base::UTF8ToUTF16(component8.c_str(), component8.size(), converted);
-#endif
-}
+}  // namespace
 
-base::string16 GetSuggestedFilenameImpl(
+// This section intends to implement the algorithm suggested in:
+// https://html.spec.whatwg.org/multipage/semantics.html#as-a-download
+// Any deviations from the spec are explicitly mentioned. The returned filename
+// is UTF-8 encoded.
+//
+// The function is named *Unsafe because the returned name isn't sanitized and
+// may contain illegal or malicious content. It should always be passed through
+// EnsureSafePortableFileName() or EnsureSafeFilenameInternal().
+//
+// TODO(asanka): Resolve deviations from the spec. The current deviations are
+// due to document origin and download attribute suggested name not being
+// reliably available at this abstraction layer. This should ideally be moved to
+// /content/.
+std::string SelectUnsafeDownloadFilename(
     const GURL& url,
     const std::string& content_disposition,
     const std::string& referrer_charset,
     const std::string& suggested_name,
     const std::string& mime_type,
     const std::string& default_name,
-    ReplaceIllegalCharactersCallback replace_illegal_characters_callback) {
-  // TODO: this function to be updated to match the httpbis recommendations.
-  // Talk to abarth for the latest news.
-
-  // We don't translate this fallback string, "download". If localization is
-  // needed, the caller should provide localized fallback in |default_name|.
-  static const base::FilePath::CharType kFinalFallbackName[] =
-      FILE_PATH_LITERAL("download");
-  std::string filename;  // In UTF-8
-  bool overwrite_extension = false;
-  bool is_name_from_content_disposition = false;
-  // Try to extract a filename from content-disposition first.
+    ExtensionGenerationOption* extension_option) {
+  *extension_option = EXTENSION_OPTION_KEEP_EXISTING;
+  // Content-Disposition comes first:
+  //
+  // Deviates from spec: The filename should only be used if the disposition
+  // type is "attachment" or if this is a trusted download as defined in section
+  // 4.6.4 of the HTML specification linked above. Currently whether this is a
+  // trusted download is not known.
   if (!content_disposition.empty()) {
     HttpContentDisposition header(content_disposition, referrer_charset);
-    filename = header.filename();
-    if (!filename.empty())
-      is_name_from_content_disposition = true;
+    if (!header.filename().empty())
+      return header.filename();
   }
 
-  // Then try to use the suggested name.
-  if (filename.empty() && !suggested_name.empty())
-    filename = suggested_name;
-
-  // Now try extracting the filename from the URL.  GetFileNameFromURL() only
+  // suggested_name is assumed to be the suggested filename from the 'download'
+  // attribute of an anchor element. It should only be non-empty if the download
+  // URL is same-origin with the interface.
+  //
+  // Deviates from spec: The filename suggested with a download attribute should
+  // also be used if the Content-Disposition specified a disposition of
+  // 'attachment' but didn't specify a filename. Currently this step can't be
+  // implemented due to suggested_name being empty if the download is not
+  // same-origin with the document.
+  if (!suggested_name.empty())
+    return suggested_name;
+
+  // Now try extracting the filename from the URL. GetFileNameFromURL() only
   // looks at the last component of the URL and doesn't return the hostname as a
   // failover.
-  if (filename.empty())
-    filename = GetFileNameFromURL(url, referrer_charset, &overwrite_extension);
+  //
+  // Deviates from spec: Extraction of a filename from the URL should only be
+  // done if the download is trusted.
+  std::string filename =
+      GetFileNameFromURL(url, referrer_charset, extension_option);
+  if (!filename.empty())
+    return filename;
+
+  // If non-empty, we can use the default before trying to extract a filename
+  // from the hostname. The latter is a last resort and is almost never what the
+  // user wanted.
+  if (!default_name.empty()) {
+    *extension_option = EXTENSION_OPTION_GENERATE_IF_MISSING;
+    return default_name;
+  }
 
   // Finally try the URL hostname, but only if there's no default specified in
   // |default_name|.  Some schemes (e.g.: file:, about:, data:) do not have a
   // host name.
-  if (filename.empty() && default_name.empty() && url.is_valid() &&
-      !url.host().empty()) {
-    // TODO(jungshik) : Decode a 'punycoded' IDN hostname. (bug 1264451)
-    filename = url.host();
+  if (url.is_valid() && !url.host().empty()) {
+    *extension_option = EXTENSION_OPTION_GENERATE_AND_APPEND;
+    base::string16 filename_string16 = IDNToUnicode(url.host(), std::string());
+    return UTF16ToUTF8(filename_string16);
   }
 
-  bool replace_trailing = false;
-  base::FilePath::StringType result_str, default_name_str;
-#if defined(OS_WIN)
-  replace_trailing = true;
-  result_str = base::UTF8ToUTF16(filename);
-  default_name_str = base::UTF8ToUTF16(default_name);
-#else
-  result_str = filename;
-  default_name_str = default_name;
-#endif
-  SanitizeGeneratedFileName(&result_str, replace_trailing);
-  if (result_str.find_last_not_of(FILE_PATH_LITERAL("-_")) ==
-      base::FilePath::StringType::npos) {
-    result_str = !default_name_str.empty()
-                     ? default_name_str
-                     : base::FilePath::StringType(kFinalFallbackName);
-    overwrite_extension = false;
-  }
-  replace_illegal_characters_callback.Run(&result_str, '-');
-  base::FilePath result(result_str);
-  // extension should not appended to filename derived from
-  // content-disposition, if it does not have one.
-  // Hence mimetype and overwrite_extension values are not used.
-  if (is_name_from_content_disposition)
-    GenerateSafeFileName("", false, &result);
-  else
-    GenerateSafeFileName(mime_type, overwrite_extension, &result);
-
-  base::string16 result16;
-  if (!FilePathToString16(result, &result16)) {
-    result = base::FilePath(default_name_str);
-    if (!FilePathToString16(result, &result16)) {
-      result = base::FilePath(kFinalFallbackName);
-      FilePathToString16(result, &result16);
-    }
-  }
-  return result16;
+  return std::string();
 }
 
-base::FilePath GenerateFileNameImpl(
-    const GURL& url,
-    const std::string& content_disposition,
-    const std::string& referrer_charset,
-    const std::string& suggested_name,
-    const std::string& mime_type,
-    const std::string& default_file_name,
-    ReplaceIllegalCharactersCallback replace_illegal_characters_callback) {
-  base::string16 file_name =
-      GetSuggestedFilenameImpl(url,
-                               content_disposition,
-                               referrer_charset,
-                               suggested_name,
-                               mime_type,
-                               default_file_name,
-                               replace_illegal_characters_callback);
+// Examines |file_name| and modifies it if necessary in order to ensure the
+// filename is safe. A new extension will be generated if |extension_option|
+// suggests so and |mime_type| is known.
+//
+// We're addressing three things here:
+//
+// 1) Usability.  If there is no reliable file extension, we want to guess a
+//    reasonable file extension based on the content type.
+//
+// 2) Shell integration.  Some file extensions automatically integrate with the
+//    shell.  We block these extensions to prevent a malicious web site from
+//    integrating with the user's shell.
+//
+// 3) Reserved names. If the resulting filename is considered reserved, then the
+//    replacement character will be inserted in front so that the file behaves
+//    as a regualr file.
+//
+// Note that this function does NOT replace illegal characters in the filename.
+// That should be done prior to calling this function.
+void EnsureSafeFilenameInternal(const std::string& mime_type,
+                                ExtensionGenerationOption extension_option,
+                                base::FilePath* file_name) {
+  base::FilePath::StringType extension = file_name->Extension();
+  if (!extension.empty())
+    extension.erase(extension.begin());  // Erase preceding '.'.
 
-#if defined(OS_WIN)
-  base::FilePath generated_name(file_name);
-#else
-  base::FilePath generated_name(
-      base::SysWideToNativeMB(base::UTF16ToWide(file_name)));
-#endif
+  switch (extension_option) {
+    case EXTENSION_OPTION_KEEP_EXISTING:
+      break;
+
+    case EXTENSION_OPTION_GENERATE_IF_MISSING:
+      if (!extension.empty())
+        break;
+    // Fallthrough
+
+    case EXTENSION_OPTION_GENERATE:
+    case EXTENSION_OPTION_GENERATE_AND_APPEND:
+      if (mime_type.empty())
+        break;
+
+      std::vector<base::FilePath::StringType> all_extensions;
+      GetExtensionsForMimeType(mime_type, &all_extensions);
+      if (all_extensions.empty())
+        // We don't expect there to be a preferred MIME extension if
+        // all_extensions was empty. The preferred extension is always expected
+        // to be among those returned by GetExtensionsForMimeType().
+        break;
+
+      if (std::find(all_extensions.begin(), all_extensions.end(), extension) !=
+          all_extensions.end())
+        break;
+      base::FilePath::StringType final_extension = file_name->FinalExtension();
+      if (!final_extension.empty()) {
+        final_extension.erase(final_extension.begin());
+        if (std::find(all_extensions.begin(), all_extensions.end(),
+                      final_extension) != all_extensions.end())
+          break;
+      }
+
+      base::FilePath::StringType new_extension;
+      GetPreferredExtensionForMimeType(mime_type, &new_extension);
+
+      // It is possible for there to be no preferred extension. In this case we
+      // leave the extension as-is. For example: application/octet-stream maps
+      // to 'exe','com' and 'bin', but none of these are preferred if there is
+      // no extension.
+      if (new_extension.empty())
+        break;
+
+      if (extension_option == EXTENSION_OPTION_GENERATE_AND_APPEND &&
+          !extension.empty()) {
+        extension.append(1, base::FilePath::kExtensionSeparator);
+        extension.append(new_extension);
+      } else {
+        extension = new_extension;
+      }
+  }
 
-  DCHECK(!generated_name.empty());
+  if (IsShellIntegratedExtension(extension))
+    extension.assign(kDefaultExtension);
 
-  return generated_name;
+  *file_name = file_name->ReplaceExtension(extension);
+
+  if (IsReservedName(file_name->value())) {
+    base::FilePath::StringType new_filename(file_name->value());
+    new_filename.insert(new_filename.begin(), 1, kReplacement);
+    *file_name = base::FilePath(new_filename);
+  }
 }
 
 }  // namespace net