in reply to where is sv_setrv?

According to newSVrv, it's 

SV_CHECK_THINKFIRST_COW_DROP(rv);
(void)SvAMAGIC_off(rv);

if (SvTYPE(rv) >= SVt_PVMG) {
    const U32 refcnt = SvREFCNT(rv);
    SvREFCNT(rv) = 0;
    sv_clear(rv);
    SvFLAGS(rv) = 0;
    SvREFCNT(rv) = refcnt;

    sv_upgrade(rv, SVt_IV);
} else if (SvROK(rv)) {
    SvREFCNT_dec(SvRV(rv));
} else if (SvTYPE(rv) < SVt_PV && SvTYPE(rv) != SVt_IV) {
    sv_upgrade(rv, SVt_IV);
} else if (SvTYPE(rv) >= SVt_PV) {
    SvPV_free(rv);
    SvLEN_set(rv, 0);
    SvCUR_set(rv, 0);
}

SvOK_off(rv);
SvRV_set(rv, sv);
SvROK_on(rv);

SvREFCNT_inc(sv);

SvSETMAGIC(rv);

[download]

Or you could use 

sv_replace(rv, newRV_inc(sv))
SvSETMAGIC(rv);

[download]

Update: Fixed ref count as per reply.

Replies are listed 'Best First'.
Re^2: where is sv_setrv?
by patcat88 (Deacon) on Dec 17, 2011 at 16:55 UTC
    Your version of newSVrv wouldn't work. Its only 5.12 compatible, not before, because of the SV type changes and the prepare_SV_for_RV macro changed. Can RVs have a PV string pointer (flags off of course)? Does an RV always have to be of type SVt_RV (5.10) or SVt_IV (5.12) or any SV with ROK can be a reference? The sv_replace seems like the safest idea while i'm writing this.

    There is a missing refcount inc on sv before "SvRV_set(rv, sv);". Also there is a macro for the old RV sv type to IV, seems to me that it will provide backwards compatibility sv.h#l75 in perl.git
[reply]

      Your version of newSVrv wouldn't work [before Perl 5.12]

      It's not a version of newSVrv, and you're free to do the exercise yourself for 5.10.

      Can RVs have a PV string pointer

      Like you just said, there's no such thing as an RV anymore.

      Does an RV always have to be of type SVt_IV (5.12) or any SV with ROK can be a reference?

      Latter. Even before 5.12.

      The sv_replace seems like the safest idea while i'm writing this.

      That's quite an understatement.

      There is a missing refcount inc on sv before "SvRV_set(rv, sv);".

      You're saying that newSVrv causes referenced material to be destroyed permaturely. I don't believe you.

[reply]
[d/l]
        There is a missing refcount inc on sv before "SvRV_set(rv, sv);".

        You're saying that newSVrv causes referenced material to be destroyed permaturely. I don't believe you.

        In newSVrv, a new target SV is created at sv.c#l9268 in perl.git. In a hypothetical setsv_rv you need to up the refcount of the EXISTING target SV that the reference will point to or the existing target SV will get garbage collected prematurely, and we arent making a weak reference here, or your following hv_store standard where the caller has to up the refcount (or do nothing with newSV) of the sv before passing to the func?
[reply]

In Section Seekers of Perl Wisdom