Remove Pickle::WriteSize() now that it has no remaining callers.

base/pickle.h

Index: base/pickle.h
===================================================================
--- base/pickle.h	(revision 125321)
+++ base/pickle.h	(working copy)
@@ -98,9 +98,15 @@
   bool WriteInt(int value) {
     return WriteBytes(&value, sizeof(value));
   }
+  // WARNING: DO NOT USE WriteLong() IF PICKLES ARE PERSISTED IN ANY WAY.
+  // WriteLong() writes whatever a "long" is on this architecture. On 32-bit
+  // platforms, it is 32 bits. On 64-bit platforms, it is 64 bits. If persisted
+  // pickles are still around after upgrading to 64-bit, or if they are copied
+  // between dissimilar systems, YOUR PICKLES WILL HAVE GONE BAD.

[jar (doing other things), 2012/03/09 22:13:36]

IMO, you have discovered a problem.... and it is a bad problem for anyone that
persists data using pickles.  

Rather than having a comment (resetting the mouse-trap, so that it will catch
the next victim, and cost us all), how abut crafting some code to handle this
issue more directly.

Can we get rid of this API?  How many users are there?  Can you migrate them to
a more stable API??

Jim

[Mike Mammarella, 2012/03/10 00:24:42]

Looks like WriteSize() might be feasible to remove, as it's not used in too many
places. (As long as all such places don't persist data, we can just change
them.) WriteLong() is used in the IPC code, which we should be able to change
safely, but it's all over the place. I'll look into it.

That said, this warning can only help compared to doing nothing, and I don't
work full-time on Chrome, so it could take a while for me to investigate this.
I'd like to add the warning for now.

   bool WriteLong(long value) {
     return WriteBytes(&value, sizeof(value));
   }
+  // WARNING: SEE THE WARNING FOR WriteLong() ABOVE. USE WriteSize() CAREFULLY.
   bool WriteSize(size_t value) {
     return WriteBytes(&value, sizeof(value));
   }