Re: Representation clauses for base-64 encoding

["Randy Brukardt" <randy@rrsoftware.com>]

"Niklas Holsti" <niklas.holsti@tidorum.invalid> wrote in message 
news:9lqpq5Ft05U1@mid.individual.net...
> On 11-12-23 03:30 , Randy Brukardt wrote:
>> "Niklas Holsti"<niklas.holsti@tidorum.invalid>  wrote in message
>> news:9lgi3jFhaU1@mid.individual.net...
>>> On 11-12-22 11:41 , Natasha Kerensikova wrote:
>> ...
>>>>      type Octet_Block is array (1 .. 3) of Octet;
>>>>      pragma Pack (Octet_Block);
>>>
>>> I would add the following, to check that packing is effective:
>>>
>>>        for Octet_Block'Size use 24;
>>
>> I think it would probably be better to forget Pack and actually say what 
>> you
>> mean here:
>>
>>      type Octet_Block is array (1 .. 3) of Octet;
>>      for Octet_Block'Component_Size use 8;
>>
>> "Pack" should be reserved for cases where you don't care about the exact
>> layout, you just want to save space. Here you do care about the exact
>> layout, so use a Component_Size.
>
> Fine adjustment: when Component_Size is not a factor or multiple of the 
> word size, RM 13.3(73) says that the array may also have to be packed in 
> order to eliminate gaps between components.
>
> Of course, a Component_Size of 8 bits is usually a factor of the word 
> size. But perhaps not always.

You're right of course, although I would hope that no compiler was silly 
enough to ignore a direct component-size command by putting gaps into it. 
I'd much rather the compiler rejected such a clause (surely Janus/Ada 
would). If you want gaps, you should say so. For instance, on the U2200 (a 
36-bit machine with 9-bit divisions), you'd have to say

     type Octet_Block is array (1 .. 3) of Octet;
     for Octet_Block'Component_Size use 9;

if you wanted gaps. (And 8 might have been rejected, I don't remember 
anymore whether the Unisys code generator could handle that.) That's why 
Janus/Ada uses a constant Host.Stor_Unit_Size in all of its representation 
clauses. (As you might guess, package Host contains a bunch of constants and 
subprograms defining the host environment for the compiler. You'd probably 
also guess that there is a similar package Target, and if you did, you'd be 
right.)

                                                     Randy.