If Perl assume all UVs were really pointers to SVs, it would crash left and right.

Remember, I'm only suggesting this can happen during cloning, not normal operations. Remember that SvRVs & SvIVs used to be different types. Maybe the cloning code, written before this amalgamation, only checked for SvTYPE and has never been updated to further distinguish RVs from IV/UVs?

Maybe you have a better explanation for this observed behaviour:

...
package main;
use threads;
use Devel::Peek;

my $o = O->new();
print $o;

$o->set( "abcde" );
print $o->get();

print "\nthreaded\n";

async {
    print $o->get( );
    $o->set( 12345 );
    print $o->get( );
}->join;

print $o->get();

__END__
C:\test>xso
N:rv:000000000035E128 o:0000000003FFA818
O=SCALAR(0x3ffa818)
S:rv;000000000035E248 o:0000000003C318F0
G:rv;000000000035E248 o:0000000003C318F0 old:000000000035E128
abcde

threaded

G:rv;0000000004258CA8 o:0000000004229090 old:0000000000000000

S:rv;0000000004258CA8 o:0000000004229090
G:rv;0000000004258CA8 o:0000000004229090 old:0000000004259920
12345
G:rv;000000000035E248 o:0000000003C318F0 old:000000000035E128
abcde
[download]

Before the thread, I set a value into the struct, and get it back.

Then I move into the thread, the RV gets cloned, the SvUV it points to gets cloned; but the value within the SvUV (labelled 'o:' in the printfs), also changes. And as the numeric value inside the UV (which we know is a pointer but Perl shouldn't), has also changed. It no longer points to the original struct. Hence, 'old:' no longer has a value.

I then insert a different value into the struct, and successfully retrieve that value. I then exit the thread and, and get the value again, and lo, I get the original value set, not the one I modified inside the thread.

No, the *only* explanation that I can come up with for that behaviour is that when the RV and UV get cloned, it also clones the thing that the UV points to; even though it shouldn't know that the numberic value in the UV is a pointer. Do you have a better explanation?

The full code:

#! perl -slw
use strict;
package O;
use Inline C => Config => BUILD_NOISY => 1;
use Inline C => <<'END_C',  NAME => 'xso', CLEAN_AFTER_BUILD => 0;

typedef struct {
    SV**sv;
} O;


void *mem( size_t size ) {
    void *p;
    Newx( p, size, char );
    return p;
}

SV *new( char *package ) {
    O *o = (O*)mem( sizeof( O ) );
    SV *rv = newRV( (SV*)o );

    o->sv = mem( sizeof( SV* ) );

    sv_bless( rv, gv_stashpv( package, 0 ) );
    SvREADONLY_on( rv );
    printf( "N:rv:%p o:%p\n", rv, o );
    return rv;
}

int set( SV *rv, SV *in ) {
    O *o = ( *(O**)SvUV( rv ) ) - 1;
    printf( "S:rv;%p o:%p\n", rv, o );
    o->sv = newSVsv( in );
    return 1;
}

SV *get( SV *rv ) {
    O *o = ( *(O**)SvUV( rv ) ) - 1;
    SV *old = newSVsv( o->sv );

    SvREFCNT_inc( old );

    printf( "G:rv;%p o:%p old:%p\n", rv, o, o->sv );
    return old;
}


END_C
package main;
use threads;
use Devel::Peek;

my $o = O->new();
print $o;

$o->set( "abcde" );
print $o->get();

print "\nthreaded\n";

async {
    print $o->get( );
    $o->set( 12345 );
    print $o->get( );
}->join;

print $o->get();
[download]

Then I move into the thread, the RV gets cloned, the SvUV it points to gets cloned; but the value within the SvUV (labelled 'o:' in the printfs), also changes.

huh? The RV doesn't point to a UV. You never even create a scalar with a UV.

Sorry. I posted the wrong version of the code. I have tried so many variations.

---

Examine what is said, not who speaks -- Silence betokens consent -- Love the truth but pardon error.

"Science is about questioning the status quo. Questioning authority".

In the absence of evidence, opinion is indistinguishable from prejudice.