Richard Leach

Before I merge this, anyone got any thoughts on how doing BODYLESS_NVs in `Perl_sv_setsv_flags` requires preservation of `SVs_TEMP`, but the BODYLESS_IV code doesn't give a hoot? Is it just because...

I looked again and think the BODYLESS_IV code in `Perl_sv_setsv_flags` _should_ do `SvOK_off`, rather than clearing **all** the existing flags. I'll modify this PR so that happens. `Perl_sv_setiv` & `Perl_sv_setnv`...

I've done the above - make `sv_setsv_flags` preserve flags like `SVs_TEMP` in the fast cases. @demerphq - do you want another look over before I merge? For posterity, what breaks...

@demerphq - do you want a final re-look through this before I merge?

Unless anyone has objections, I'd like to merge this before the next dev release in just over a week's time.

No further comments received on this one since Yves said it looked okay, so am merging.

Stepping through `pp_subst` with the script above: * `TARG` is never `SvIsCOW` at the start of `pp_subst` (`was_cow` == 0) * `force_on_match` == 0 * `r_flags` == 1 * `CALLREGEXEC(rx,...

Possibly fixed with: ``` diff --git a/lib/B/Deparse.pm b/lib/B/Deparse.pm index 3d1d8b36cc..3ce1f28d4e 100644 --- a/lib/B/Deparse.pm +++ b/lib/B/Deparse.pm @@ -5129,7 +5129,7 @@ sub retscalar { |msgrcv|semop|semget|semctl|hintseval|shostent|snetent |sprotoent|sservent|ehostent|enetent|eprotoent|eservent |spwent|epwent|sgrent|egrent|getlogin|syscall|lock|runcv - |fc)\z/x + |fc|padsv_store)\z/x }...

This is an alternative to https://github.com/Perl/perl5/pull/20318

I noticed the other day that in the _scalar_refs_ case, the assignment is wrapped in a pp_once, whereas that's not the case for the _declared_refs_ case. The problem might be...