Use factory to create histograms, and refcounts to track lifetimes...

net/base/mime_sniffer.cc

 // Copyright (c) 2006-2008 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 // Detecting mime types is a tricky business because we need to balance
 // compatibility concerns with security issues.  Here is a survey of how other
 // browsers behave and then a description of how we intend to behave.
 //
 // HTML payload, no Content-Type header:
 // * IE 7: Render as HTML
@@ skipping 81 matching lines @@
 
 #include "net/base/mime_sniffer.h"
 
 #include "base/basictypes.h"
 #include "base/histogram.h"
 #include "base/logging.h"
 #include "base/string_util.h"
 #include "googleurl/src/gurl.h"
 #include "net/base/mime_util.h"
 
-namespace {
-
-class SnifferHistogram : public LinearHistogram {
- public:
-  SnifferHistogram(const char* name, int array_size)
-      : LinearHistogram(name, 0, array_size - 1, array_size) {
-    SetFlags(kUmaTargetedHistogramFlag);
-  }
-};
-
-}  // namespace
-
 namespace net {
 
 // We aren't interested in looking at more than 512 bytes of content
 static const size_t kMaxBytesToSniff = 512;
 
 // The number of content bytes we need to use all our magic numbers.  Feel free
 // to increase this number if you add a longer magic number.
 static const size_t kBytesRequiredForMagic = 42;
 
 struct MagicNumber {
@@ skipping 87 matching lines @@
   MAGIC_HTML_TAG("table")  // Mozilla
   MAGIC_HTML_TAG("a")  // Mozilla
   MAGIC_HTML_TAG("style")  // Mozilla
   MAGIC_HTML_TAG("title")  // Mozilla
   MAGIC_HTML_TAG("b")  // Mozilla
   MAGIC_HTML_TAG("body")  // Mozilla
   MAGIC_HTML_TAG("br")
   MAGIC_HTML_TAG("p")  // Mozilla
 };
 
+static scoped_refptr<Histogram> UMASnifferHistogramGet(const char* name,
+                                                       int array_size) {
+  scoped_refptr<Histogram> counter =
+      LinearHistogram::LinearHistogramFactoryGet(
+          name, 1, array_size - 1, array_size);
+  counter->SetFlags(kUmaTargetedHistogramFlag);
+  return counter;
+}
+
 static bool MatchMagicNumber(const char* content, size_t size,
                              const MagicNumber* magic_entry,
                              std::string* result) {
   const size_t len = magic_entry->magic_len;
 
   // Keep kBytesRequiredForMagic honest.
   DCHECK(len <= kBytesRequiredForMagic);
 
   // To compare with magic strings, we need to compute strlen(content), but
   // content might not actually have a null terminator.  In that case, we
@@ skipping 35 matching lines @@
 static bool SniffForHTML(const char* content, size_t size,
                          std::string* result) {
   // We adopt a strategy similar to that used by Mozilla to sniff HTML tags,
   // but with some modifications to better match the HTML5 spec.
   const char* const end = content + size;
   const char* pos;
   for (pos = content; pos < end; ++pos) {
     if (!IsAsciiWhitespace(*pos))
       break;
   }
-  static SnifferHistogram counter("mime_sniffer.kSniffableTags2",
-                                  arraysize(kSniffableTags));
+  static scoped_refptr<Histogram> counter =
+      UMASnifferHistogramGet("mime_sniffer.kSniffableTags2",
+                             arraysize(kSniffableTags));
   // |pos| now points to first non-whitespace character (or at end).
   return CheckForMagicNumbers(pos, end - pos,
                               kSniffableTags, arraysize(kSniffableTags),
-                              &counter, result);
+                              counter.get(), result);
 }
 
 static bool SniffForMagicNumbers(const char* content, size_t size,
                                  std::string* result) {
   // Check our big table of Magic Numbers
-  static SnifferHistogram counter("mime_sniffer.kMagicNumbers2",
-                                  arraysize(kMagicNumbers));
+  static scoped_refptr<Histogram> counter =
+      UMASnifferHistogramGet("mime_sniffer.kMagicNumbers2",
+                             arraysize(kMagicNumbers));
   return CheckForMagicNumbers(content, size,
                               kMagicNumbers, arraysize(kMagicNumbers),
-                              &counter, result);
+                              counter.get(), result);
 }
 
 // Byte order marks
 static const MagicNumber kMagicXML[] = {
   // We want to be very conservative in interpreting text/xml content as
   // XHTML -- we just want to sniff enough to make unit tests pass.
   // So we match explicitly on this, and don't match other ways of writing
   // it in semantically-equivalent ways.
   MAGIC_STRING("application/xhtml+xml",
                "<html xmlns=\"http://www.w3.org/1999/xhtml\"")
@@ skipping 11 matching lines @@
   // We allow at most kFirstTagBytes bytes of content before we expect the
   // opening tag.
   const size_t kFeedAllowedHeaderBytes = 300;
   const char* const end = content + std::min(size, kFeedAllowedHeaderBytes);
   const char* pos = content;
 
   // This loop iterates through tag-looking offsets in the file.
   // We want to skip XML processing instructions (of the form "<?xml ...")
   // and stop at the first "plain" tag, then make a decision on the mime-type
   // based on the name (or possibly attributes) of that tag.
-  static SnifferHistogram counter("mime_sniffer.kMagicXML2",
-                                  arraysize(kMagicXML));
+  static scoped_refptr<Histogram> counter =
+      UMASnifferHistogramGet("mime_sniffer.kMagicXML2",
+                             arraysize(kMagicXML));
   const int kMaxTagIterations = 5;
   for (int i = 0; i < kMaxTagIterations && pos < end; ++i) {
     pos = reinterpret_cast<const char*>(memchr(pos, '<', end - pos));
     if (!pos)
       return false;
 
     if (base::strncasecmp(pos, "<?xml", sizeof("<?xml")-1) == 0) {
       // Skip XML declarations.
       ++pos;
       continue;
     } else if (base::strncasecmp(pos, "<!DOCTYPE",
                                  sizeof("<!DOCTYPE")-1) == 0) {
       // Skip DOCTYPE declarations.
       ++pos;
       continue;
     }
 
     if (CheckForMagicNumbers(pos, end - pos,
                              kMagicXML, arraysize(kMagicXML),
-                             &counter, result))
+                             counter.get(), result))
       return true;
 
     // TODO(evanm): handle RSS 1.0, which is an RDF format and more difficult
     // to identify.
 
     // If we get here, we've hit an initial tag that hasn't matched one of the
     // above tests.  Abort.
     return true;
   }
 
@@ skipping 26 matching lines @@
   0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,  // 0xA0 - 0xAF
   0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,  // 0xB0 - 0xBF
   0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,  // 0xC0 - 0xCF
   0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,  // 0xD0 - 0xDF
   0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,  // 0xE0 - 0xEF
   0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,  // 0xF0 - 0xFF
 };
 
 static bool LooksBinary(const char* content, size_t size) {
   // First, we look for a BOM.
-  static SnifferHistogram counter("mime_sniffer.kByteOrderMark2",
-                                  arraysize(kByteOrderMark));
+  static scoped_refptr<Histogram> counter =
+      UMASnifferHistogramGet("mime_sniffer.kByteOrderMark2",
+                             arraysize(kByteOrderMark));
   std::string unused;
   if (CheckForMagicNumbers(content, size,
                            kByteOrderMark, arraysize(kByteOrderMark),
-                           &counter, &unused)) {
+                           counter.get(), &unused)) {
     // If there is BOM, we think the buffer is not binary.
     return false;
   }
 
   // Next we look to see if any of the bytes "look binary."
   for (size_t i = 0; i < size; ++i) {
     // If we a see a binary-looking byte, we think the content is binary.
     if (kByteLooksBinary[static_cast<unsigned char>(content[i])])
       return true;
   }
 
   // No evidence either way, default to non-binary.
   return false;
 }
 
 static bool IsUnknownMimeType(const std::string& mime_type) {
   // TODO(tc): Maybe reuse some code in net/http/http_response_headers.* here.
   // If we do, please be careful not to alter the semantics at all.
   static const char* kUnknownMimeTypes[] = {
     // Empty mime types are as unknown as they get.
     "",
     // The unknown/unknown type is popular and uninformative
     "unknown/unknown",
     // The second most popular unknown mime type is application/unknown
     "application/unknown",
     // Firefox rejects a mime type if it is exactly */*
     "*/*",
   };
-  static SnifferHistogram counter("mime_sniffer.kUnknownMimeTypes2",
-                                  arraysize(kUnknownMimeTypes) + 1);
+  static scoped_refptr<Histogram> counter =
+      UMASnifferHistogramGet("mime_sniffer.kUnknownMimeTypes2",
+                             arraysize(kUnknownMimeTypes) + 1);
   for (size_t i = 0; i < arraysize(kUnknownMimeTypes); ++i) {
     if (mime_type == kUnknownMimeTypes[i]) {
-      counter.Add(i);
+      counter->Add(i);
       return true;
     }
   }
   if (mime_type.find('/') == std::string::npos) {
     // Firefox rejects a mime type if it does not contain a slash
-    counter.Add(arraysize(kUnknownMimeTypes));
+    counter->Add(arraysize(kUnknownMimeTypes));
     return true;
   }
   return false;
 }
 
 // Sniff a crx (chrome extension) file.
 static bool SniffCRX(const char* content, size_t content_size, const GURL& url,
                      const std::string& type_hint, std::string* result) {
-  static SnifferHistogram counter("mime_sniffer.kSniffCRX", 3);
+  static scoped_refptr<Histogram> counter =
+      UMASnifferHistogramGet("mime_sniffer.kSniffCRX", 3);
 
   // Technically, the crx magic number is just Cr24, but the bytes after that
   // are a version number which changes infrequently. Including it in the
   // sniffing gives us less room for error. If the version number ever changes,
   // we can just add an entry to this list.
   //
   // TODO(aa): If we ever have another magic number, we'll want to pass a
   // histogram into CheckForMagicNumbers(), below, to see which one matched.
   const struct MagicNumber kCRXMagicNumbers[] = {
     MAGIC_NUMBER("application/x-chrome-extension", "Cr24\x02\x00\x00\x00")
   };
 
   // Only consider files that have the extension ".crx".
   const char kCRXExtension[] = ".crx";
   const int kExtensionLength = arraysize(kCRXExtension) - 1;  // ignore null
   if (url.path().rfind(kCRXExtension, std::string::npos, kExtensionLength) ==
       url.path().size() - kExtensionLength) {
-    counter.Add(1);
+    counter->Add(1);
   } else {
     return false;
   }
 
   if (CheckForMagicNumbers(content, content_size,
                            kCRXMagicNumbers, arraysize(kCRXMagicNumbers),
                            NULL, result)) {
-    counter.Add(2);
+    counter->Add(2);
   } else {
     return false;
   }
 
   return true;
 }
 
 bool ShouldSniffMimeType(const GURL& url, const std::string& mime_type) {
-  static SnifferHistogram should_sniff_counter(
-      "mime_sniffer.ShouldSniffMimeType2", 3);
+  static scoped_refptr<Histogram> should_sniff_counter =
+      UMASnifferHistogramGet("mime_sniffer.ShouldSniffMimeType2", 3);
   // We are willing to sniff the mime type for HTTP, HTTPS, and FTP
   bool sniffable_scheme = url.is_empty() ||
                           url.SchemeIs("http") ||
                           url.SchemeIs("https") ||
                           url.SchemeIs("ftp");
   if (!sniffable_scheme) {
-    should_sniff_counter.Add(1);
+    should_sniff_counter->Add(1);
     return false;
   }
 
   static const char* kSniffableTypes[] = {
     // Many web servers are misconfigured to send text/plain for many
     // different types of content.
     "text/plain",
     // IIS 4.0 and 5.0 send application/octet-stream when serving .xhtml
     // files.  Firefox 2.0 does not sniff xhtml here, but Safari 3,
     // Opera 9, and IE do.
     "application/octet-stream",
     // XHTML and Atom/RSS feeds are often served as plain xml instead of
     // their more specific mime types.
     "text/xml",
     "application/xml",
   };
-  static SnifferHistogram counter("mime_sniffer.kSniffableTypes2",
-                                  arraysize(kSniffableTypes) + 1);
+  static scoped_refptr<Histogram> counter =
+      UMASnifferHistogramGet("mime_sniffer.kSniffableTypes2",
+                             arraysize(kSniffableTypes) + 1);
   for (size_t i = 0; i < arraysize(kSniffableTypes); ++i) {
     if (mime_type == kSniffableTypes[i]) {
-      counter.Add(i);
-      should_sniff_counter.Add(2);
+      counter->Add(i);
+      should_sniff_counter->Add(2);
       return true;
     }
   }
   if (IsUnknownMimeType(mime_type)) {
     // The web server didn't specify a content type or specified a mime
     // type that we ignore.
-    counter.Add(arraysize(kSniffableTypes));
-    should_sniff_counter.Add(2);
+    counter->Add(arraysize(kSniffableTypes));
+    should_sniff_counter->Add(2);
     return true;
   }
-  should_sniff_counter.Add(1);
+  should_sniff_counter->Add(1);
   return false;
 }
 
 bool SniffMimeType(const char* content, size_t content_size,
                    const GURL& url, const std::string& type_hint,
                    std::string* result) {
   DCHECK_LT(content_size, 1000000U);  // sanity check
   DCHECK(content);
   DCHECK(result);
 
@@ skipping 60 matching lines @@
       result->assign("text/plain");
     // We could change our mind if a binary-looking byte appears later in
     // the content, so we only have enough content if we have the max.
     return content_size >= kMaxBytesToSniff;
   }
 
   return have_enough_content;
 }
 
 }  // namespace net