[Info-vax] problem with 64-bit pointers in C

[Jan-Erik Soderholm]

Den 2018-01-31 kl. 13:40, skrev Joukj:
> John Reagan wrote:
>> The largest possible variable is 32-bits big even with 64-bit pointers.
>>
>> You can only subtract pointers that point inside the same variable.
>> Therefore they can never be more than 32-bits apart. 
> 
> So this will not bite me in the application : Although both pointer are 
> 64-bit the difference result should be 32-bit.
> But still you always get the warning.....
> 
>>
>> That's why ptrdiff_t is always 32.  The type is built into the frontend 
>> and can't easily be tweaked by header changes.
>>
>> You notice that size_t is also 32 even with 64-bit pointers.
>>
>> Looking back, better choices could have been made.  Going forward, I'd 
>> like to have a real LP64 mode with 'long' being 64 to match compilers 
>> like clang.
>>
>> For now, typecast one of them to __int64 at the subtract.
> When casting one of them I get a a compiler error
>     printf( "%lld\n" , (__int64)buffer - entry_start );
> ......................^
> %CC-E-NOSUBTRACT, In this statement, "entry_start" cannot be subtracted 
> from "(6
> 4 bit int)buffer".
> at line number 13 in file $DISK6:[JOUKJ.test]test.c;86
> 
> Type casting both is just the same as I did in the last line of the test 
> and gives you sizeof(int)*<expected result>
> 

This "works" (with a warning):

    printf( "%d  %d  %d\n" , sizeof( entry_start ) ,
            sizeof( (__int64)(entry_start - buffer) ) ,
            sizeof( buffer ) );

It typecasts the *result* of the subtraction, not the individual
variables. Now, it can still not be a values larger than 32 bits,
I guess, even if it is later typecasted to 64 bits...)




> 
> The reason for asking this question is that I tried to compile the pnglib 
> package with 64 bit pointers. It compiled with some warnings and the 
> self-test failed. So I'm trying to reduce the compile warnings first.
> 
> 
>                         regards
>                             Jouk