[Lazarus] INVALID_* consts

[Hans-Peter Diettrich]

Simple solution:

The INVALID_FILE_SIZE constant is used only inside Win32 RTL code, with 
API records or API functions returning exactly that type and value. This 
constant is not exported from the unit. Win32 specific user code, 
missing this constant value, has to be updated.

The user only gets platform-independent int64 filesizes, and here -1 or 
every negative value can indicate an error.

DoDi


waldo kitty schrieb:
> On 6/3/2013 18:54, Hans-Peter Diettrich wrote:
>> waldo kitty schrieb:
>>> what can we do about INVALID_FILE_SIZE?
>>>
>>> const
>>> INVALID_FILE_SIZE = DWORD($FFFFFFFF);
>>>
>>> is the above correct? it would seem to be from what i can make out 
>>> from MSDN
>>> and the various C code files i've found it defined in...
>>
>> Using unsigned constants with int64 filesizes would classify files of 
>> exactly
>> that size as invalid, while smaller or bigger files would be valid. This
>> certainly is *not* what the coder had in mind, so that I'd suggest to 
>> inspect
>> and fix all occurences of this (and similar) unsigned constants in 
>> user code.
> 
> i'm (finally) back to trying to poke this conversion thing with a sharp 
> stick... i understand what you are saying in the above but i don't know 
> how to "fix all occurrences of this (and similar) unsigned constants in 
> user code."... i'm sure it is probably something easy that i'm not 
> seeing... something like adding a sign ;)
> 
>   eg: INVALID_FILE_SIZE = DWORD(-1);
> 
> but how can i know if that will be what OS API functions send back? the 
> above const is documented as shown above in the DWORD form so i already 
> know that that particular instance from that particular API function 
> will fail... since DWORD($FFFFFFFF) <> -1...
> 
>>> i kind of have to laugh because on place i read spoke of examples 
>>> checking the
>>> result directly instead of using INVALID_FILE_SIZE and that this was 
>>> not good
>>> coding practice because m$ might change the value of 
>>> INVALID_FILE_SIZE at any
>>> time which would then break all of that code... i makes sense to me O:)
>>
>> IMO such unsigned constants should be used only in specific cases, 
>> e.g. with OS
>> defined records containing part of an longer (int64) type. The FCL can 
>> use that
>> constant in Win32 platform code, the LCL or user code never should use 
>> it at all.
> 
> i understand this... how can i get this const (as defined) into the FCL 
> if it does not already exist? how can i tell my converted delphi project 
> to use the FCL in addition to the LCL? right now when i look at the 
> project inspector, the LCL is the only "required" package shown...
> 
>> I also doubt that the MS documentation is up to date whenever above 
>> constant is
>> used with anything except struct fields of type DWORD. Even in 32 bit 
>> code
>> *both* low/high parts of a 64 bit filesize must be checked for
>> INVALID_FILE_SIZE, in order to determine a valid result.
> 
> the documentation for the above const (which i linked to previously) 
> talks about using a different OS API function for those instances where 
> necessary...
> 
> 
> -- 
> _______________________________________________
> Lazarus mailing list
> Lazarus at lists.lazarus.freepascal.org
> http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus
>