Re: First time Ada has let me down

["Ken Garlington" <Ken.Garlington@computer.org>]

"Shifty" <avlogue@home.com> wrote in message
news:8FD7DEBEEsynoptikdamudderfuck@news...
> OK, here's the story:
>
> I had a problem at work where I needed to calculate the offset of a
> "magic number" within a buffer containing an IP datagram.
>
> The magic number is 3 protocol headers deep within the datagram, and
> all three headers are variable length.  I needed to find the HLEN
> fields of the IP and TCP headers in memory.  Turns out both of these
> are stored in 4 bits.  The data in the other 4 bits of the octet
> belong to another field in the header.
>
> This stumped me for awhile, but then I went home and picked up a C book.
> Hmmm, chapter 15 - bit fiddling.  I learned a bunch of stuff about masks
> and bitwise operators and came up with a working solution.  (which seems
> much better than using record representation clauses and typecasting a 4
> bit value into an 8 bit integer)

I'm curious about the "typecasting a 4 bit value into an 8 bit integer"
statement. Why would you need to do this?

I would have thought the record representation clause approach would be much
preferable than a bunch of
#DEFINEs to describe the header layout. Granted, the fact that the headers
are variable length complicates
things a bit, but just looking at an IPv4 header, I would expect something
like the source code below. In fact,
if you use this approach (records with variant), you might be able to avoid
some of the offset calculations altogether.
You certainly don't need to do any low-level bit twiddling (which is why
this wasn't as big of a deal in Ada83 as
you might think...)

package IPv4 is

type Byte is range 0 .. 16#FF#;
for Byte'Size use 8;

type Version_Type is range 4 .. 4;
subtype Header_Length_Type is Integer range 5 .. 15;
subtype Total_Length_Type is Integer range 0 .. 65_535;

-- more types for the rest of the header

type Option_Type is array (1 .. 4) of Byte;
pragma Pack (Option_Type);
type Option_Type_Array is array (Integer range <>) of Option_Type;
pragma Pack (Option_Type_Array);

type Header_Type (IHL : Header_Length_Type) is record
   Version : Version_Type := 4;
   Type_of_Service : Byte := 0;
   Total_Length : Total_Length_Type;

   -- the rest of the mandatory fields

   case IHL is
      when 5 =>
         null;
      when 6 .. Header_Length_Type'Last =>
         Option : Option_Type_Array(6 .. IHL);
   end case;
end record;

-- record rep spec goes here

end IPv4;

> Next morning I tried to implement the solution and couldn't find the Ada
> equivalent of C's bit-wise & operater.  According to my "Ada as a 2nd
> language" book, the Ada reserved word "AND" only works for boolean types,
> not integer types.  I got the impression that it was up to the programmer
> to write his own bitwise ANDs/ORs/XORs, etc (it wouldn't be hard).
>
> Luckily I found a vendor-supplied package which provided this
> functionality, but I can't believe Ada (83) doesn't have native
> language support for these!!!  Please tell me that I am hopelessly
> confused and dead wrong!
>
> Cheers,
> Andrew.