is atexit() availabel for fpc 3.2.2 Application's ?

[paule32]

Hi,
is there a Windows win32api Function that act like atexit() from the LibC-Library ?
So, when Halt() is called, the atexit set Routine is called.

[Fibonacci]

https://www.freepascal.org/docs-html/rtl/system/exitproc.html

https://wiki.freepascal.org/ExitProc

Not implemented in the mini RTL, yet

[paule32]

Thanks fibonacci.
never see, not yet testet a procedure end;

[Fibonacci]

Code: Pascal  [Select][+][-]

1. program atexit;
2.  
3. procedure x;
4. begin
5.   writeln('xxxx');
6.   readln;
7. end;
8.  
9. begin
10.   ExitProc := @x;
11.   writeln('1');
12.   Halt;
13.   writeln('2');
14.   readln;
15. end.

[Thaddy]

Better is
https://www.freepascal.org/docs-html/rtl/system/addexitproc.html

Code: Pascal  [Select][+][-]

1. {$mode objfpc}
2. program demoexit;
3. procedure helloexit;
4. begin
5.   writeln('Going to the exit');
6. end;
7.  
8. begin
9.   AddExitProc(@helloexit);
10. end.

The reason is that this keeps the exitproc chain intact, whereas the previous example does not.
That can cause issues.
The last added in the chain is executed last.

@Fibonacci
You destroyed that chain. Although it is a "working" example it is wrong.
There may be frameworks that rely on that chain to release resources or otherwise.
The correct way is as per my example.
I know that the documentation lacks a bit here, both the wiki - which is plain wrong - and the real docs.
It comes from TurboPascal:
"Compared to the simpler ExitProc, AddExitProc automatically preserves the existing chain, ensuring that previously registered exit procedures are not overwritten. This is particularly useful in larger applications or frameworks that rely on proper resource cleanup."

[Thaddy]

Ergo: the TP docs are better here.
When used correctly it is very powerful.

[Fibonacci]

Nah its not wrong. There is no chain if you dont create one, nothing is destroyed. You should know Paule is coding weird stuff, right? In this case Paule wants the ExitProc, trust me, I know the context.

[Thaddy]

Yes, he is trying to code way beyond his current capabilities. Which is admirable, but needs corrections.
In my opinion it is wrong to directly assign to exitproc, though.
For the reasons I mentioned.

I will correct the wiki entry this weekend. It is mindblowingly wrong.