[gdal-dev] RFC 31 - OGR 64bit Fields and FID

[Joaquim Luis]

On 10-09-2010 18:50, Even Rouault wrote:
> Frank,
>
> yes I definitely think that 64bit integer field/FID support is a must-have for
> OGR.
>
> Here are my remarks on the proposal :
>
> 0) Minor point : about "Note that the old interfaces using "long" are already
> 64bit on 64bit operating systems so there is little harm to applications
> continuing to use these interfaces on 64bit operating systems". This is true
> on almost all platforms, except on Win64 where a long is still 32 bit...
>    

Hi,

About the issue of the 64-bits ints on Windows we had quite some 
troubles in GMT while implementing it.
The problem, as far as I remember and I'm the one who suffered more with 
it (other developers use Macs only), was that one cannot read as %ld a 
variable that is a 64 bits int. There was no compiling or running error, 
but memory was corrupted at totally unexpected/unrelated places.
We have a note in one of our headers that describe it, and the solution 
we adopted

/* Note: Under Windows 64-bit a 64-bit integer is __int64 and when used
  * with scanf the format must be %lld.  This is not exactly what we call
  * POSIX-clean where %ld is expected.  Thus, in places where such 64-bit
  * variables are processed we let the compiler build the actual format
  * using the GMT_LL string which is either "l" or "ll"
  */
#ifdef _WIN64
typedef __int64 GMT_LONG;    /* A signed 8-byte integer under 64-bit 
Windows */
#define GMT_LL "ll"
#else
typedef long GMT_LONG;        /* A signed 4 (or 8-byte for 64-bit) 
integer */
#define GMT_LL "l"
#endif


than we use it for example like

     sscanf (p, "%" GMT_LL "d-%" GMT_LL "d", &start, &stop);


Excuse if this trivia for you guys (it wasn't for us)

Joaquim Luis