linux-traipu team mailing list archive

[Stewart Smith <stewart@xxxxxxxxxxxxxxxx>]

** Changed in: drizzle
   Importance: Undecided => Medium

** Changed in: drizzle
       Status: New => Triaged

-- 
You received this bug notification because you are a member of UBUNTU -
AL - BR, which is subscribed to Drizzle.
https://bugs.launchpad.net/bugs/1131492

Title:
  JS function can't handle ints > 32 bits.

Status in A Lightweight SQL Database for Cloud Infrastructure and Web Applications:
  Triaged

Bug description:
  try:

  select js('arguments[0]', 1<<31),  js('arguments[0]', 1<<32);

  gives:

  +---------------------------+---------------------------+
  | js('arguments[0]', 1<<31) | js('arguments[0]', 1<<32) |
  +---------------------------+---------------------------+
  | 2147483648              | 0                                  |
  +---------------------------+---------------------------+

  Expected was:

  +---------------------------+---------------------------+
  | js('arguments[0]', 1<<31) | js('arguments[0]', 1<<32) |
  +---------------------------+---------------------------+
  | 2147483648                | 4294967296                |
  +---------------------------+---------------------------+

  The problem seems to be that js uses v8::Integer for all arguments of arg_type INTEGER_RESULT. 
  It should probably something like:

  long long int_val = *((long long *)args->args[i]);
  if (int_val >> 31) return v8::Number::New((double)int_val);
  else return v8::Integer::New(int_val);

  logic: 
  INTEGER_RESULT means a 64bit integer. 
  v8::Integer can only hold 32 bit integer. 
  INTEGER_RESULT >> 31 would truncate all bits of a 32 bit integer save for the sign bit. 
  If that sign bit is set, we would be dealing with a positive int in MySQL land which would become a negative int in v8 land. 
  And if any of the more significant bits would be set, then we would simply have an integer that cannot fit in a 32 bit integer at all.  
  So, if INTEGER_RESULT >> 31 is true (one or more bits set) it is out of range for a v8::Integer, and it should be represented by a v8::Number. In order to do that we should cast the long long value to a double since that's what the v8::Number wants to receive.

To manage notifications about this bug go to:
https://bugs.launchpad.net/drizzle/+bug/1131492/+subscriptions