const declaration

[Warfley]

ChatGPT says that c = GetC; is ok, but Lazarus doesn't argree, it says

This is because large language models are not really good at being correct. They often hallucinate (the actual academic term) things which look correct and are great for producing a lot of stuff thats mostly correct. But they are not reliable in questions of details.

So while you can ask ChatGPT to generate you some code, or to explain to you something, it will only be like 80% correct. To efficiently use AI for programming you must know more about programming than the AI model. You can then delegate some of the tasks to the AI model, but must be prepared to spend some time correcting it.

[silvestre]

Warfley, what you describe is exactly how the situation is today and some people are not very aware of it. AI has yet to evolve to understand the “small” details that make the differences, and programmers, need to continue to have a solid foundation in coding to use AI correctly.

ChatGPT says that c = GetC; is ok, but Lazarus doesn't argree, it says

This is because large language models are not really good at being correct. They often hallucinate (the actual academic term) things which look correct and are great for producing a lot of stuff thats mostly correct. But they are not reliable in questions of details.

So while you can ask ChatGPT to generate you some code, or to explain to you something, it will only be like 80% correct. To efficiently use AI for programming you must know more about programming than the AI model. You can then delegate some of the tasks to the AI model, but must be prepared to spend some time correcting it.

[Sieben]

Code: Pascal  [Select][+][-]

1. program project1;
2.  
3. {$mode objfpc}{$H+}
4.  
5. uses
6.   SysUtils;
7.  
8. function GetC(): integer;
9. begin
10.   Result := 1 + 2;
11. end;
12.  
13. const
14.   a = 1;
15.   b = a + 1;
16.   c = GetC();
17.  
18. begin
19.   writeln(inttostr(a));
20.   writeln(inttostr(b));
21.   writeln(inttostr(c));
22. end.    
23.  

Where's the pleasure in that...? Why would you want to do that? Why not just:

Code: Pascal  [Select][+][-]

1. const
2.   a = 1;
3.   b = a + 1;
4.   c = 3;

and in any scenario when you would call GetC() just use c...?

[Joanna from IRC]

Why are discussions about chatgpt and AI still happening here ?
 It’s completely off topic and has nothing to do with fpc .

It should be obvious to anyone who bothers to learn pascal that a function result is not a constant. 

[Thaddy]

It is not off- topic when people actually use it and chatGPT comes up with an answer that we have to correct. By now I have instructed chatGPT to not use functions in variable declarations with default values.(yes, you can do that). Takes some time but it will use that knowledge eventually.

[Joanna from IRC]

Why on earth would someone rather use chatgpt instead of learning to code properly??  and moreover it is not our responsibility to act as unpaid volunteers to help improve the product of a for profit company that wants to put all creators of intellectual property out of business!

Nobody should be encouraged to use this AI garbage at all.

[Zvoni]

Where's the pleasure in that...? Why would you want to do that? Why not just:

Code: Pascal  [Select][+][-]

1. const
2.   a = 1;
3.   b = a + 1;
4.   c = 3;

and in any scenario when you would call GetC() just use c...?

Because in C you can pass values for defined Symbols during compilation, so TS labours under the misconception that it's possible in FPC, and that those passed values are THEN used during Compilation, to "calculate" the value for the constant

[Thaddy]

More in general, if chatGPT makes a mistake, point that out and show the code that is wrong to chatGPT.
The model will adapt itself.

[MarkMLl]

More in general, if chatGPT makes a mistake, point that out and show the code that is wrong to chatGPT.
The model will adapt itself.

But other models, which by now are using this forum as an authoritative reference, will not.

MarkMLl

[Thaddy]

Mark, it will eventually sort itself out.

[Martin_fr]

Mark, it will eventually sort itself out.

Hopefully... And in the sense "sort out" meaning "remove from availability"? I.e. the LLMs will (remove) sort themself out from the list of useful tools.

If they do, that may be their first smart move.

[Joanna from IRC]

People are infinitely gullible when it comes to trying to get something for nothing...

[freemind001]

"Why would anybody ever DO that???"

i'm trying to make a wrapper for iunput.h in pascal for a personal project.

there are things like

Code: C  [Select][+][-]

1. /* ioctl */
2. #define UINPUT_IOCTL_BASE       'U'
3. #define UI_DEV_CREATE           _IO(UINPUT_IOCTL_BASE, 1)
4. #define UI_DEV_DESTROY          _IO(UINPUT_IOCTL_BASE, 2)

which seem to be not implementable in pascal, as you can't pass a value to macros.

As the project is for fun, talking about salary doesn't make sense. As well as discussing the knowledge of Pascal of forum members.
And yes, I agree, ChatGPT is pretty stupid when dealing with pascal.

[Thaddy]

They are perfectly implementable, but you need to implement them as inlined functions.
That is no different from the C implementation, it is only that C macro's allow for parameters, but in effect they end up as a function anyway.

Also note that C macro's are notoriously unsafe!, no type checking is done.
I think that was one of the reasons to not expand Freepascal's macro capabilities.

Just write the macro as a function.
For a real programmer that is easy. Very easy.

[MarkMLl]

which seem to be not implementable in pascal, as you can't pass a value to macros.

Looks as though it's related to this that I did for talking over SPI:

Code: Pascal  [Select][+][-]

1. function _IOC(dir, _type, nr: Longword; size: SizeInt): TIOCtlRequest;
2.  
3. begin
4. {$push }{$R- } // If dir is 2 (_IOC_READ) this will set the MSB hence a range error
5.   _IOC := (dir  shl _IOC_DIRSHIFT) or
6.           (_type shl _IOC_TYPESHIFT) or
7.           (nr   shl _IOC_NRSHIFT) or
8.           (size shl _IOC_SIZESHIFT);
9. {$pop }
10. end { _IOC } ;
11.  
12.  
13. function _IOR(_type, nr: Cardinal; size: SizeInt): TIOCtlRequest;
14.  
15. begin
16.   _IOR := _IOC(_IOC_READ, _type, nr, size);
17. end { _ IOR } ;
18.  
19.  
20. function _IOW(_type, nr: cardinal; size: SizeInt): TIOCtlRequest;
21.  
22. begin
23.   _IOW := _IOC(_IOC_WRITE, _type, nr, size);
24. end { _IOW } ;
25.  
26.  
27. function SPI_MSGSIZE(n: SizeInt): SizeInt; inline;
28.  
29. begin
30.   if (n * sizeof(spi_ioc_transfer)) < (1 shl _IOC_SIZEBITS) then
31.     SPI_MSGSIZE := n * sizeof(spi_ioc_transfer)
32.   else
33.     SPI_MSGSIZE := 0;
34. end { SPI_MSGSIZE } ;
35.  
36.  
37. function SPI_IOC_MESSAGE(N: SizeInt): Cardinal; inline;
38.  
39. begin
40.   SPI_IOC_MESSAGE := _IOW(SPI_IOC_MAGIC, 0, SPI_MSGSIZE(N));
41. end { SPI_IOC_MESSAGE } ;
42.  

So, not really any problem at all (provided that you're absolutely rigorous with parameter sizes) even if there I've not inlined the functions.

/However/, I refuse to be an apologist for Pascal's lack of parameterised macros, which are even- IIUC- implemented by Rust.

MarkMLl