Re: storage error: stack overflow

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

On 08/13/2015 04:19 AM, Simon Wright wrote:
> hreba <f_hreba@yahoo.com.br> writes:
>
>> ==4823== Invalid read of size 4
>> ==4823==    at 0x66FF1A0: ada__strings__unbounded__reference (in
>> /usr/lib/x86_64-linux-gnu/libgnat-4.6.so.1)
>> ==4823==    by 0x6703851: ada__strings__unbounded___assign__2 (in
>> /usr/lib/x86_64-linux-gnu/libgnat-4.6.so.1)
>
> There's no Assign in the code of Ada.Strings.Unbounded, so I expect it's
> a subprogram generated by the compiler to manage assignment.
>
> It could be from the finalize of the 'dest' object prior to assignment,
> but I think it's from the adjust after the bitwise copy of the 'source'
> object; Adjust does contain a call to Reference.
>
> Given that, is it possible that something has corrupted the 'source'
> object? (at least 'source.name')
>

The statement
	
     dest.name := source.name;

is the first line in the body of the subprogram

    procedure Deep_Copy_Base
       (source: Surface; dest: in out Surface'Class);

Setting a breakpoint at this subprogram, "p source" gets me

$1 = (name => (prev => 0x0, next => 0x0, reference => 0x0), dist => 0.0, 
zabs => 1500.0, zmin => 0.0, zmax => 0.0, ap => 0x722a90)

and on the other Linux machine, without the error, I get

$1 = (name => (prev => 0x0, next => 0x0, reference => 0xf22b98, last => 
0), dist => 0.0, zabs => 1500.0, zmin => 0.0, zmax => 0.0,
     ap => 0x8126b80)

The value of source.name comes from a call of 'To_Unbounded_String("")'.
The null reference is strange.

I made a simple test program with 2 variables a, b of type 
'Unbounded_String' and an assignment and it worked. After assigning the 
empty string to a its value is

$1 = (prev => 0x0, next => 0x0, reference => 
0x7ffff7dca6a0)<ada.strings.unbounded.empty_shared_string>)

I reinstalled GNAT, GPS and GtkAda from the Ubuntu repository and 
nothing changed. I'll continue my investigation when I have more time.

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