Merge branch 'webp-update'

1 files changed, 119 insertions, 0 deletions

diff --git a/image/FrameTimeout.h b/image/FrameTimeout.h
new file mode 100644
index 000000000..4070bba65
--- /dev/null
+++ b/image/FrameTimeout.h
@@ -0,0 +1,119 @@
+/* -*- Mode: C++; tab-width: 2; indent-tabs-mode: nil; c-basic-offset: 2 -*-
+ *
+ * This Source Code Form is subject to the terms of the Mozilla Public
+ * License, v. 2.0. If a copy of the MPL was not distributed with this
+ * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
+
+#ifndef mozilla_image_FrameTimeout_h
+#define mozilla_image_FrameTimeout_h
+
+#include <stdint.h>
+#include "mozilla/Assertions.h"
+
+namespace mozilla {
+namespace image {
+
+/**
+ * FrameTimeout wraps a frame timeout value (measured in milliseconds) after
+ * first normalizing it. This normalization is necessary because some tools
+ * generate incorrect frame timeout values which we nevertheless have to
+ * support. For this reason, code that deals with frame timeouts should always
+ * use a FrameTimeout value rather than the raw value from the image header.
+ */
+struct FrameTimeout
+{
+  /**
+   * @return a FrameTimeout of zero. This should be used only for math
+   * involving FrameTimeout values. You can't obtain a zero FrameTimeout from
+   * FromRawMilliseconds().
+   */
+  static FrameTimeout Zero() { return FrameTimeout(0); }
+
+  /// @return an infinite FrameTimeout.
+  static FrameTimeout Forever() { return FrameTimeout(-1); }
+
+  /// @return a FrameTimeout obtained by normalizing a raw timeout value.
+  static FrameTimeout FromRawMilliseconds(int32_t aRawMilliseconds)
+  {
+    // Normalize all infinite timeouts to the same value.
+    if (aRawMilliseconds < 0) {
+      return FrameTimeout::Forever();
+    }
+
+    // Very small timeout values are problematic for two reasons: we don't want
+    // to burn energy redrawing animated images extremely fast, and broken tools
+    // generate these values when they actually want a "default" value, so such
+    // images won't play back right without normalization. For some context,
+    // see bug 890743, bug 125137, bug 139677, and bug 207059. The historical
+    // behavior of IE and Opera was:
+    //   IE 6/Win:
+    //     10 - 50ms is normalized to 100ms.
+    //     >50ms is used unnormalized.
+    //   Opera 7 final/Win:
+    //     10ms is normalized to 100ms.
+    //     >10ms is used unnormalized.
+    if (aRawMilliseconds >= 0 && aRawMilliseconds <= 10 ) {
+      return FrameTimeout(100);
+    }
+
+    // The provided timeout value is OK as-is.
+    return FrameTimeout(aRawMilliseconds);
+  }
+
+  bool operator==(const FrameTimeout& aOther) const
+  {
+    return mTimeout == aOther.mTimeout;
+  }
+
+  bool operator!=(const FrameTimeout& aOther) const { return !(*this == aOther); }
+
+  FrameTimeout operator+(const FrameTimeout& aOther)
+  {
+    if (*this == Forever() || aOther == Forever()) {
+      return Forever();
+    }
+
+    return FrameTimeout(mTimeout + aOther.mTimeout);
+  }
+
+  FrameTimeout& operator+=(const FrameTimeout& aOther)
+  {
+    *this = *this + aOther;
+    return *this;
+  }
+
+  /**
+   * @return this FrameTimeout's value in milliseconds. Illegal to call on a
+   * an infinite FrameTimeout value.
+   */
+  uint32_t AsMilliseconds() const
+  {
+    if (*this == Forever()) {
+      MOZ_ASSERT_UNREACHABLE("Calling AsMilliseconds() on an infinite FrameTimeout");
+      return 100;  // Fail to something sane.
+    }
+
+    return uint32_t(mTimeout);
+  }
+
+  /**
+   * @return this FrameTimeout value encoded so that non-negative values
+   * represent a timeout in milliseconds, and -1 represents an infinite
+   * timeout.
+   *
+   * XXX(seth): This is a backwards compatibility hack that should be removed.
+   */
+  int32_t AsEncodedValueDeprecated() const { return mTimeout; }
+
+private:
+  explicit FrameTimeout(int32_t aTimeout)
+    : mTimeout(aTimeout)
+  { }
+
+  int32_t mTimeout;
+};
+
+} // namespace image
+} // namespace mozilla
+
+#endif // mozilla_image_FrameTimeout_h