Re^4: Perl XS portable uint32

Ah this just demonstrates the bit loss. Here is an example using 2 different width ints:

use Inline C => <<'EOC';

void poo () {
    unsigned short s = 0xffff;
    unsigned long  l = (unsigned long) s;
    printf("short %08x\n long %08x\n", sizeof(s), sizeof(l));
    printf("short %08x\n long %08x\n", s, l);
    s <<= 4; l <<= 4;
    printf("short %08x\n long %08x\n", s, l);
    s >>= 4; l >>= 4;
    printf("short %08x\n long %08x\n", s, l);
    if ( l != (unsigned long) s ) 
        printf("Bugger!\nerror %08x\n", l^(unsigned long)s );
}

EOC

poo();

__END__
short 00000002
 long 00000004
short 0000ffff
 long 0000ffff
short 0000fff0
 long 000ffff0
short 00000fff
 long 0000ffff
Bugger!
error 0000f000
[download]

This is exactly what happens if you think you have a 32 bit int but actually have a 64 bit int. 2 operations and the results now differ.....

Comment on Re^4: Perl XS portable uint32_t Download Code

Replies are listed 'Best First'.
Re^5: Perl XS portable uint32_t by syphilis (Archbishop) on Jun 06, 2008 at 22:30 UTC
This is exactly what happens if you think you have a 32 bit int but actually have a 64 bit int Yes - which demonstrates that you need to know what you have. I'm in agreement with that - and have been right from the beginning. It's just that I don't see a need to establish anything other than that wrt Digest::JHash. And, given that that info can easily be derived by (or passed to) the pre-processor, you can tailor jhash() to efficiently deal with any scenario. But maybe I'm being blind to certain issues ... so I'll just butt out. I look forward to the next release of Digest::JHash, where I get to see how you eventually do deal with the issue :-) Cheers, Rob	[reply]

Replies are listed 'Best First'.

Re^5: Perl XS portable uint32_t
by syphilis (Archbishop) on Jun 06, 2008 at 22:30 UTC

This is exactly what happens if you think you have a 32 bit int but actually have a 64 bit int

[reply]