Front page | perl.perl5.porters |
Postings from November 2010
On Mon, Nov 15, 2010 at 11:52:18PM +0100, Father Chrysostomos wrote:
> commit 2d0d1eccfcfe45f2fac66080cd30117a608088f8
> Author: Father Chrysostomos <sprout@cpan.org>
> Date: Mon Nov 15 14:30:07 2010 -0800
>
> [perl #79208] %stash:: = () anonymises CVs
>
> This keeps stash names visible during %foo:: = ().
>
> This fixes @ISA assignment inside a DESTROY method triggered by
> %foo:: = () and also lets existing CVs retain their pointers to
> the stash.
>
> So
> %foo:: = ()
> is now equivalent to
> delete $foo::{$_} for keys %foo::
>
> M hv.c
> M t/op/stash.t
> if (SvOOK(hv)) {
> HE *entry;
> struct mro_meta *meta;
> - struct xpvhv_aux *iter = HvAUX(hv);
> +
> + SvFLAGS(hv) &= ~SVf_OOK; /* Goodbye, aux structure. */
> + /* What aux structure? */
> + /* (But we still have a pointer to it in iter.) */
> +
> + /* Copy the name to a new aux structure if there is a name. */
> + if (iter->xhv_name) {
> + struct xpvhv_aux * const newaux = hv_auxinit(hv);
^^^^^^^^^
> + newaux->xhv_name = iter->xhv_name;
> + newaux->xhv_name_count = iter->xhv_name_count;
> + iter->xhv_name = NULL;
> + }
> +
> + /* Because we have taken xhv_name out, the only allocated
> + pointers in the aux structure that might exist are the back-
> + reference array, xhv_eiter and the MRO stuff.
> + */
> +
That leaks in some cases. I'm not sure what the correct fix should be. (Sorry)
$ PERL_DESTRUCT_LEVEL=2 valgrind --leak-check=full ./miniperl -Ilib configpm
==20144== Memcheck, a memory error detector
==20144== Copyright (C) 2002-2010, and GNU GPL'd, by Julian Seward et al.
==20144== Using Valgrind-3.6.0 and LibVEX; rerun with -h for copyright info
==20144== Command: ./miniperl -Ilib configpm
==20144==
written lib/Config.pod
==20144== Warning: bad signal number 0 in sigaction()
==20144==
==20144== HEAP SUMMARY:
==20144== in use at exit: 1,888 bytes in 16 blocks
==20144== total heap usage: 40,267 allocs, 40,251 frees, 7,278,070 bytes allocated
==20144==
==20144== 1,888 (164 direct, 1,724 indirect) bytes in 1 blocks are definitely lost in loss record 4 of 4
==20144== at 0x40235C4: malloc (vg_replace_malloc.c:236)
==20144== by 0x81099AE: Perl_safesyscalloc (util.c:321)
==20144== by 0x8141ED7: S_hv_auxinit (hv.c:1919)
==20144== by 0x8140E07: S_hfreeentries (hv.c:1683)
==20144== by 0x81418E8: Perl_hv_undef (hv.c:1842)
==20144== by 0x81AAE1D: Perl_sv_clear (sv.c:6051)
==20144== by 0x81AC2FA: Perl_sv_free2 (sv.c:6328)
==20144== by 0x805B6D5: Perl_gp_free (gv.c:1698)
==20144== by 0x81AB44D: Perl_sv_clear (sv.c:6085)
==20144== by 0x81AC2FA: Perl_sv_free2 (sv.c:6328)
==20144== by 0x81702D2: do_clean_all (sv.c:589)
==20144== by 0x816EA23: S_visit (sv.c:411)
==20144==
==20144== LEAK SUMMARY:
==20144== definitely lost: 164 bytes in 1 blocks
==20144== indirectly lost: 1,724 bytes in 15 blocks
==20144== possibly lost: 0 bytes in 0 blocks
==20144== still reachable: 0 bytes in 0 blocks
==20144== suppressed: 0 bytes in 0 blocks
==20144==
==20144== For counts of detected and suppressed errors, rerun with: -v
==20144== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 23 from 6)
I suspect "except during global destruction" is not the correct, general-case
fix.
Nicholas Clark
Thread Previous