[fpc-devel] Attn Joost: the ampersand in dwarf for var-param

[Martin]

On 14/01/2011 22:16, Joost van der Sluis wrote:
> On Fri, 2011-01-14 at 21:27 +0000, Martin wrote:
>> Been to quick...
>> Found the answer
> I though you'd noticed this earlier. I also wrote it in one of the mails
> last week?
Seems I have missed that, sorry.
Anyway, that is good news...

>> But it seems that now, even in dwarf-3  objects are treated as pointer
>> again? (I can happily live with that. I just want to know)
> Yes, that was the (final?) conclusion from the discussion with you and
> Jonas. It is also why I said you were too fast with your tests, because
> I still had to change this.
That is also good news.

Especially since I found a better way to distinguish between
var
   a: TObject;
   b: ^TObject;
in order to display the correct data in the IDE.
And the new way (once implemented) will also work if the debug info is 
mixed dwarf/stabs

ptype a^
ptype b^

will return correct (note the ^ in front of TObject)
a^:  ~"type = TObject = class : ....
b^:  ~"type = ^TObject = class : .....

And that is the same for all kind of debug infos :)

And with luck, that will work with the Mac 6.3.50 as well (because 
6.3.50 cannot do "whatis TObject" => and current IDE wants that...)



> Note that there will be more changes. Now effectively the Dwarf-2 code
> for writing object-info is used for Dwarf 3 also. But there were some
> additions in them old Dwarf-3 code, that I have to re-implement. (In
> such a way that as much as possible code is shared between Dwarf-2 and
> Dwarf-3)
I will start filling in reports...

Have you sen the RunGdbmi app next to my test? It allows you to run a 
set of gbd instructions to the same application, with different 
debuggers[1], different compilers[1], and different gw/gs settings, and 
see all the result at once.

[1] assuming you use gdblist.txt and fpclist.txt files

The advantage is, you see the actual gdb output, instead of what the IDE 
does with it....

Martin