NEWSV'd scalar isn't initially empty?

diotalevi has asked for the wisdom of the Perl Monks concerning the following question:

I wrote the following Inline::C function to work through what it takes to create a scalar successfully. The string portion of the scalar is not null and I thought it should be. Am I doing this wrong or am I expected to zero it out manually?

$ perl new.pl
SV = PV(0x7460) at 0x35cb4
  REFCNT = 1
  FLAGS = (TEMP)
  PV = 0x6330 " \0w.pl\0\0\0\0"\0
  CUR = 10
  LEN = 11
[download]

use Devel::Peek;
use Inline C=> 'DATA';

Dump(foo( 10 ));

sub vis { map { join '', map sprintf("\\%o", ord), split // } @_ }

__DATA__
__C__
SV* foo(int nr) {
    SV* out = NEWSV(535,nr);
    SvUPGRADE(out,SVt_PV);
    SvCUR_set(out,nr);
    return (out);
}
[download]

Comment on NEWSV'd scalar isn't initially empty? Select or Download Code

Replies are listed 'Best First'.
Re: NEWSV'd scalar isn't initially empty? by diotalevi (Canon) on Aug 04, 2003 at 16:09 UTC
Ah. On reading the perl source I think I should have written this thusly. This leaves off the upgrade and grow (that's apparently handled within NEWSV when len > 0) and since NEWSV uses New() instead of Newz(), I had to follow up with an explicit Zero() on the PV. So... is this right? I added a SvPOK_only() call as well because without that the only flag on it was TEMP. If I do `my $foo = foo(); Dump($foo)` the REFCNT is one - is that right and am I supposed to mortalize the scalar I'm returning here? `SV* foo(int nr) { SV* out = NEWSV(535,nr); SvPOK_only(out); Zero(SvPVX(out), nr, char); SvCUR_set(out,nr); return (out); }` [download]	[reply] [d/l] [select]
Re: Re: NEWSV'd scalar isn't initially empty? by mpeppler (Vicar) on Aug 05, 2003 at 09:01 UTC
Well - the REFCNT of `$foo` has to be one - it is a copy of the SV * returned from foo(). I am pretty sure that you should add `sv_2mortal(out)` to your function to avoid leaking the SV. Michael	[reply] [d/l] [select]