I am getting compile hints "Local variable "<variable name>" does not seem to be initialized" for a number of variables of packed record type.

Does not the FPC compiler reserve and zero the memory area of such variables?

In any case I am using Move() to fill the variables with data from a binary buffer, maybe the compiler is not smart enough to detect this and remove the warning?

(FPC 3.0 and Lazarus 1.6 on Raspberry Pi3)

No.

In any case I am using Move() to fill the variables with data from a binary buffer, maybe the compiler is not smart enough to detect this and remove the warning?

The compiler has no way of knowing if your move statement initializes the full record.

Michael.

Michael Van Canneyt wrote

> Does not the FPC compiler reserve and zero the memory area of such > variables? No.

Does thios mean that the memory is reserved but not cleared or does it mean that memory is not reserved at all?

Well, if it has no way of knowing that the record is stuffed with data then why does it issue a notice that it seems uninitialized? Are there other commands it does detect as initialization of the variable?

the memory is reserved, but not initialized.

not that I know of.

you need to assign it a value, you can assign it default(yourtype)

Michael.

OK, I did not know about the default keyword.

Is "default" zero?
i.e.

var
  HD: TSSReadingHeader;
  ..
begin
  HD := Default(TSSReadingHeader);

And now HD.anything is zero?

OK, I did not know about the default keyword. Is "default" zero?

http://wiki.lazarus.freepascal.org/FPC_New_Features_3.0#New_compiler_intrinsic_Default

Or better yet, the official documentation:

http://www.freepascal.org/docs-html/current/rtl/system/default.html

Michael.

Can the examples be added to official docs?

Mattias

Good point, will do !

Michael.

Michael Van Canneyt wrote

Or better yet, the official documentation: http://www.freepascal.org/docs-html/current/rtl/system/default.html

Question please. Using the official documentation as linked to above, how does one find out what versions of the compiler/rtl/fcl support something?

One does not - yet.

I am working on leaving the various versions available (notice the 'current' in the URL) and then you could view previous versions...

Michael.

I made a test once I got the other issues solved and indeed Default() on my packed records seems to zero out the memory! Very good to know! :)

Thanks for pointing it out!

I started hacking through trying to produce at least a list of functions/procedures and a permuted index from different versions, but I got bogged down in an apparent requirement that the documentation files needed to be processed by the corresponding version of supplied utilities. I'm sure what I did was of little value since I was attacking the problem with insufficient understanding of "the right way to do it" and I'm not even sure where it's got to.