Re: storage error: stack overflow

[hreba <f_hreba@yahoo.com.br>]

On 08/18/2015 04:24 AM, Egil H H wrote:
> On Tuesday, August 18, 2015 at 4:16:38 AM UTC+2, hreba wrote:
>>
>> with Ada.Strings.Unbounded;
>>
>> procedure testUBS is
>>
>>      type TA is record
>>            name:		Ada.Strings.Unbounded.Unbounded_String;
>>      end record;
>>
>>      type TB is record
>>         a:	TA;
>>      end record;
>>
>>      function New_TB  (name: String:="")  return access TB is
>>      begin
>>         return new TB'
>>           (a=>TA'(name=>Ada.Strings.Unbounded.To_Unbounded_String(name)));
>>      end New_TB;
>>
>>      b:	 access TB:= New_TB (name=>"Det");
>>      name: Ada.Strings.Unbounded.Unbounded_String;
>>
>> begin
>>      name:= b.a.name;
>> end testUBS;
>>
>
> This fails as described with Gnat Pro 6.4.1 (GCC 4.5.2),
> but seems to be fixed in later versions, at least Gnat Pro 7.1.1 (GCC 4.7.3).
>
> You should try to avoid anonymous access types if at all possible.

I need an access type because in my real program TB is a tagged type and 
I need class-wide variables. As long as everything works I prefer the 
anonymous type because it saves me a type conversion. But I am still in 
doubt about the best practice, there is some discrepancy between the 
examples in my Ada book (Barnes) and the posts in this group.

> However, you could try to work around the problem using an extended return,
> like this:
>
> function New_TB
>     (Name : String := "")
>      return access TB
> is
> begin
>     return Foo : access TB := new TB do
>        foo.all := TB'
>           (A => TA'
>              (Name => Ada.Strings.Unbounded.To_Unbounded_String(Name)));
>     end return;
> end New_TB;
>

Didn't know about extended return (the Ada 95 book was cheaper -:). But 
there is a problem: the TB in my original program is not only tagged, it 
is limited too (and New_TB is for initialization). So "foo.all:=..." is 
forbidden.

-- 
Frank Hrebabetzky		+55 / 48 / 3235 1106
Florianopolis, Brazil