Re: Ada.Strings.Fixed.Count raises Storage_Error

[rieachus@comcast.net]

On Thursday, June 30, 2016 at 1:38:50 AM UTC-4, Niklas Holsti wrote:

> Regarding Rosen's pathology, your explanation shows that it was a severe 
> one and really required a change in the language definition. A different 
> kind of thing than the possible problems of Strings extending to 
> Integer'Last.

No it is not.  There was a huge amount of wrangling about the consequences of the type Integer being 16-bits in size in Ada 83.  The classic issue was that code for loops generated normally would never terminate for a loop over Integer'Range.  (Strings were fine because they began at +1.)

Over the next few years, it became common for even Ada compilers for 16-bit word size systems to support a 32-bit Integer type. Then came machines with 64-bit integer registers and operations, and now, RM 3.5.4(11), Integer is a constrained subtype, and 64-bit registers can be used to count Integer loops without a problem.  But the problem does come back for 64-bit or 128-bit or whatever types.  No one is going to write a test which would outlast the physical machine, so we can ignore the issue--sort of.  You can go write a loop that runs from System.Min_Int .. System.Max_Int and inspect the code generated to see if it is correct.

I consider this issue to be a subcase of this pathology.  On machines with 64-bit registers and a 32-bit Integer type, the generated code should be fine.