[fsp] Improve performance for reading small chunks of data.

webkit/browser/fileapi/file_system_context.h

Index: webkit/browser/fileapi/file_system_context.h
diff --git a/webkit/browser/fileapi/file_system_context.h b/webkit/browser/fileapi/file_system_context.h
index 428ef601149e660da3b49cdb710d5704a26609ed..a69afb6552e7b900954e1d55381eae94ed09d111 100644
--- a/webkit/browser/fileapi/file_system_context.h
+++ b/webkit/browser/fileapi/file_system_context.h
@@ -246,9 +246,12 @@ class WEBKIT_STORAGE_BROWSER_EXPORT FileSystemContext
   // FileSystemBackend for the URL and call the backend's CreateFileReader.
   // The resolved FileSystemBackend could perform further specialization
   // depending on the filesystem type pointed by the |url|.
+  // The |length| argument says how many bytes are going to be read using the
+  // instance of the file stream reader. If unknown, then equal to -1.

[hashimoto, 2014/08/22 06:53:42]

Why don't you use kuint64max (or kint64max) to be consistent with existing code?
This way, you don't have to handle -1 specially.

[mtomasz, 2014/08/25 03:32:31]

Seems that kuint64max and kint64max are deprecated. So instead of simple -1 we
would have the following code:

#include <limits>
(...)
CreateFileStreamReader(..., numeric_limits<uint64>::max(), ...)

This is just way longer than -1. Also IMHO less readable.

[hashimoto, 2014/09/01 11:01:01]

I'm asking about consistency with the existing code, not about the length or the
readability of the code.
Also, using -1 is less intuitive than numeric_limits<uint64>::max(). (at least
-1 should be a constant.)

   scoped_ptr<webkit_blob::FileStreamReader> CreateFileStreamReader(
       const FileSystemURL& url,
       int64 offset,
+      int64 length,
       const base::Time& expected_modification_time);
 
   // Creates new FileStreamWriter instance to write into a file pointed by