Hi,

Fpc sometimes complains (hints) about local variables not being initialized. My question: how do I initialize more complicated structures like records? And should I need to (I'm guessing no in the example below, but the contents of the record will be undefined - no problem as there are no pointers/strings in there)?

Example:
....
Type
  TIfdTag = packed Record
    ID: word;
    //Tag number
    Typ: word;
    //Type tag
    Count: cardinal;
    //tag length
    Offset: cardinal;
    //Offset / Value
  End;
....
Var
  tag: TIfdTag;
....
in code somewhere
               ReadTag(tag);

As TIfdTag seems to contain non-pointer records only, I thought I could use Sizeof and tried this:
FillByte(tag, SizeOf(tag), 0); //initialize var.
The compiler warning doesn't go away though.

Is this the recommended way of doing things?

Thanks,

2010/9/27 Jim :

There is not enough code shown to say much. If the ReadTag procedure is declared with a presumably var paramater of type TIfdTag then the compiler warning is OK if the passed record has not been initialized in any way. In such case it may/should help to change the 'var' paramater to an 'out' parameter to explicitly tell the compiler that there is nothing going into ReadTag but only the other way.

HTH

It's enough code, because FillByte (and others) are defined with a "var" parameter and such the hint (not a warning) won't go away. It's an old topic already and we have just to live with it (regarding the Fill* functions). ^^

Regards,
Sven

Thanks you two, that helped!

It's enough code, because FillByte (and others) are defined with a "var" parameter and such the hint (not a warning) won't go away. It's an old topic already and we have just to live with it (regarding the Fill* functions). ^^

But why not simply change the definition of the variable to "out"? Then the Fill* functions would be recognized as initializing routines.

2010/9/28 Jürgen Hestermann :

No, see http://bugs.freepascal.org/view.php?id=17020

Vincent

@Jürgen:
That's what I meant, when I said it's an old topic. ;)

Regards,
Sven

Jim wrote

FillByte(tag, SizeOf(tag), 0); //initialize var. The compiler warning doesn't go away though.

I've had the same problem a few monts ago, and I hate compiler hints/warnings in my code. The issue is the var parameters, and the FPC developers already stated that they will not change the Fill*() functions parameters to being out parameters (some Windows OLE or COM thing). What I had to do was create my own wrapper function for Fill(), then inside that wrapper function I called FillChar or FillByte etc, and that call is wrapped by {$HINT OFF} and {$HINT ON} directives. A bit of a pain, but it worked, and I silenced the compiler warning.

It seems that we need a new parameter modifier "out_non_finalize" since "out" includes a finalization.

We had a similar discussion with "constref".

Maybe we need a comprehensive method to describe calling conventions, see e.g. http://mantis.freepascal.org/view.php?id=16697 or the discussion about "in out" parameters: http://lists.freepascal.org/lists/fpc-devel/2010-August/021452.html

GCC has some stuff like that for inline assembly, see e.g. the paragraph "Extended inline assembly" in http://www.delorie.com/djgpp/doc/brennan/brennan_att_inline_djgpp.html ___________________________________________________________ WEB.DE DSL SOMMER-SPECIAL: Surf & Phone Flat 16.000 für nur 19,99 €/mtl.!* http://produkte.web.de/go/DSL_Doppel_Flatrate/2

Yes, I see now.
Over a longer time the link above did not work for me (error 400 or so) so I could not look at it and therefore did not know that "out" is no solution.