Getting address of TSomeObject.AfterConstruction without using PVMT
lazarus / 64-FPC development

----------------------------------------------------------------------
BrunoK · 2019-02-02 09:58

In the following procedure
Code: Pascal  [Select][+][-]window.onload = function(){var x1 =
document.getElementById("main_content_section"); if (x1) { var x =
document.getElementsByClassName("geshi");for (var i = 0; i const 
cCPEmptyMethod : Codepointer = nil;  cCPAfterConstruction :
Codepointer = nil; function _InternalCreate(aSelf : TClass; aCreate:
UIntPtr): TObject;var  lAfterConstruction : CodePointer;begin  Result
:= aSelf.Newinstance;  if Result = nil then    exit;  {
TObject.AfterConstruction does nothing, skip try / finally block thus 
  reducing code overhead for 90% of classes (?) }  lAfterConstruction
:= PVMT(aSelf)^.vAfterConstruction;  // lAfterConstruction :=
@Result.AfterConstruction; // if activated, generates error  //
fstobjcr.pp(96,25) Error: Incompatible types: got "<procedure variable
type of procedure of object;Register>" expected "Pointer"  if
lAfterConstruction <> cCPAfterConstruction then    if
lAfterConstruction <> cCPEmptyMethod then     try     
 lAfterConstruction := nil;       Result.AfterConstruction;     
 lAfterConstruction := Result;     finally       if lAfterConstruction
= nil then begin // Failed AfterConstruction         Result.Destroy; 
      Result := nil;       end;     end;end; I would like to replace 
PVMT(aSelf)^.vAfterConstruction with something like shown in the next
line. But expressed  like in snippet line 15, it gives error like in
line 16.
Is there a way to get 'simply' (without having to go through
complicated things that I already know how to do) a correct
transtyping of lAfterConstruction := @Result.AfterConstruction.

----------------------------------------------------------------------
howardpc · 2019-02-02 10:57

The following compiles. Whether it achieves what you want, I am not
sure.
Code: Pascal  [Select][+][-]window.onload = function(){var x1 =
document.getElementById("main_content_section"); if (x1) { var x =
document.getElementsByClassName("geshi");for (var i = 0; i uses
Forms; function _InternalCreatedOK(aSelf: TClass; out anInstance:
TObject): Boolean;var  lAfterConstruction: TProcedureOfObject;begin 
Result := True;  anInstance := aSelf.newinstance;  if anInstance = Nil
then    Exit(False);  lAfterConstruction :=
@anInstance.AfterConstruction; // if activated, generates error  try 
 if Assigned(lAfterConstruction) then    lAfterConstruction := Nil;  
anInstance.AfterConstruction;  finally    if
Assigned(lAfterConstruction) then begin // Failed AfterConstruction  
  anInstance.Destroy;      anInstance := Nil;      Result := False;  
end;  end;end;

----------------------------------------------------------------------
ASerge · 2019-02-02 11:06

May be this:
Code: Pascal  [Select][+][-]window.onload = function(){var x1 =
document.getElementById("main_content_section"); if (x1) { var x =
document.getElementsByClassName("geshi");for (var i = 0; i {$MODE
OBJFPC} var  GAfterConstruction: CodePointer = nil; type 
TAfterContruction = procedure of object; function
GetCodeOfAfterContruction(Method: TAfterContruction): CodePointer;
inline;begin  Result := TMethod(Method).Code;end; function
_InternalCreate(ASelf: TClass; aCreate: UIntPtr): TObject;var 
LAfterConstruction: CodePointer;begin  Result := ASelf.NewInstance; 
if Result = nil then    Exit;{ TObject.AfterConstruction does nothing,
skip try / finally block thus  reducing code overhead for 90% of
classes (?) }  LAfterConstruction :=
GetCodeOfAfterContruction(@Result.AfterConstruction);  if
LAfterConstruction = GAfterConstruction then    // Skip default  else 
  Result.AfterConstruction;end; procedure InitGAfterConstruction;var 
Dummy: TObject;begin  Dummy := TObject.Create;  try   
GAfterConstruction :=
GetCodeOfAfterContruction(@Dummy.AfterConstruction);  finally   
Dummy.Free;  end;end; begin  InitGAfterConstruction;end.

----------------------------------------------------------------------
BrunoK · 2019-02-02 11:49

@howardpc, @ASerge
Thanks. As soon as my 3.0.6 is again functional (some little
modifications beiing done), I'll test.
The objective is to have as little byte code as possible, otherwise
what I'm trying to do do is useless (once again ...).