Problem with debugger
lazarus / 12-Debugger

----------------------------------------------------------------------
shadeofgray · 2007-12-20 19:29

Hello!

I am using Win32 version of Lazarus (0.9.24 beta) and have problems
with debugger. Here is sample code:

Code: [Select]
procedure TForm1.BtnTestClick(Sender: TObject);
var
    A:      array of Double;
    v:      double;
begin
    SetLength(A, 1);
    a[0]:=2;
    v:=a[0];
    Memo1.Lines.Add('DONE');
end;


I placed breakpoint after the last line. At this breakpoint I used
Evaluate/Modify tool and got "v=2" , "a[0]=2.343973527005793e-308".
And i got even stranger results when I tried to use Evaluate/Modify on
arrays which are members of data structures.

What does it mean? Is Evaluate/Modify tool unfinished, unstable or
what?

----------------------------------------------------------------------
shadeofgray · 2007-12-21 07:18

May be the right question is - does anyone experience same behavior?

----------------------------------------------------------------------
Marc · 2007-12-21 10:16

evaluating (dyn) arrays is tricky and iirc not supported.
Problem is that gdb is very C oriented, so language constructs which
are not possible in C usually cannot be evaluated.

----------------------------------------------------------------------
Martin_fr · 2008-01-03 17:26

hm I done a bit of debugging:

the problem is that a dynarray is in truth a pointer to an array.

actually it points into the middle of a structure...

  var a : array of Integer;
  setlength(a,22)

then a contins the address x

x-4 : integer = high(a) = 21
x     : frst element
x +1: 2nd element
....

if it was array of integer then the 2nd element is at x+4.
I do not know what happens if you use byte alignment?

So how to access the true value?

*** WARNING FIRST ***
The below code means you access memory through pointers and calculated
values. You could end up anywhere, and if you make a type, or use an
index out of bounds or something else on your system differs to mine
=>
you may get a crash, you may hang your entire operating system, you
may ??? well anything can happen.  BUT usually, you will just get an
error message if you mess up.
***


Well if you have a : array of integer; // or a 4byte numeri type,
then:

(a[0]+0)^
(a[0]+4)^
(a[0]+8)^

and for the size [high(a)]
(a[0]-4)^


in case of " a: array of byte" it is more fun (anybody got an easier
solution?)

^BYTE(^LONGINT(@a1)^+0)^
^BYTE(^LONGINT(@a1)^+1)^
^BYTE(^LONGINT(@a1)^+2)^
and for the size
^BYTE(^LONGINT(@a1)^-4)^

----------------

ok now my question:

IMHO this is a bug in fpc? because "ptype" should not be an
"array[0..-1] of BYTE", it should be a pointer to this datatype.

Any opinions on this?

----------------------------------------------------------------------
Marc · 2008-01-04 08:53

maybe it should, but question remains, does gdb understand ?
(if so, better ask thison the freepascal mailinglist or forum)