32 bit Type Error in MAT_v5.jl

[fundamental]

It would appear that div() was returning Int64 values in some cases which generates type errors on 32 bit machines. The below patch appears to resolve the issue as with its addition all tests pass (read.jl&write.jl).

diff --git a/src/MAT_v5.jl b/src/MAT_v5.jl
index d1a7938..a8d04f1 100644
--- a/src/MAT_v5.jl
+++ b/src/MAT_v5.jl
@@ -92,7 +92,7 @@ end

 function read_element{T}(f::IOStream, swap_bytes::Bool, ::Type{T})
     (dtype, nbytes, hbytes) = read_header(f, swap_bytes)
-    data = read_bswap(f, swap_bytes, T, div(nbytes, sizeof(T)))
+    data = read_bswap(f, swap_bytes, T, int(div(nbytes, sizeof(T))))
     skip_padding(f, nbytes, hbytes)
     data
 end
@@ -191,7 +191,7 @@ function read_string(f::IOStream, swap_bytes::Bool, dimensions::Vector{Int32})
         # Technically, if dtype == 3 or dtype == 4, this is ISO-8859-1 and not Unicode.
         # However, the first 256 Unicode code points are derived from ISO-8859-1, so UCS-2
         # is a superset of 2-byte ISO-8859-1.
-        chars = read_bswap(f, swap_bytes, Uint16, div(nbytes, 2))
+        chars = read_bswap(f, swap_bytes, Uint16, int(div(nbytes, 2)))
         if dimensions[1] == 1
             data = CharString(chars)
         else