Front page | perl.perl5.porters |
Postings from November 2003
On Wed, Oct 08, 2003 at 10:41:55PM -0400, Rick Delaney wrote:
> On Wed, Oct 08, 2003 at 10:31:01AM +0530, Abhijit Menon-Sen wrote:
> > At 2003-10-07 12:51:51 +0200, h.m.brand@hccnet.nl wrote:
> > >
> > > linux [locale:en_US]-DDEBUGGING -Accflags='-DPERL_COPY_ON_WRITE'
^^^^^^^^^^^^^^^^^^
> > > ../lib/locale.t.........................FAILED 99-117
> > > ../t/comp/require.t.....................FAILED 28-41
> >
> > I can't reproduce these failures. Could someone else please investigate?
>
> I can't reproduce them either but try the following for require.t since
> I overlooked that. Is there any advantage to making these "skip tests"?
>
> I looked at locale.t and it similarly bails after test 98 if $have_setlocale
> is false but I can't see any connection to the require patch. It would
> be nice to see the output of the test itself.
I think that the locale.t failure is due to a known bug with copy on write,
which is best summarised as:
perl -C8 -e '${^OPEN} =~ /(:utf8)/'
Assertion ((sstr)->sv_flags & 0x00040000) failed: file "sv.c", line 4202 at -e line 1.
Finally, I think I have found the cause. Hopefully change 21774 fixes it.
Nicholas Clark
Change 21774 by nicholas@entropy on 2003/11/22 21:21:22
restore_magic needs to undo COW
(while magic was off, sv_setsv is likely to have seen this SV as a
good source for COW)
Should fix the smoke failures on lib/locale.t
Affected files ...
... //depot/perl/mg.c#283 edit
Differences ...
==== //depot/perl/mg.c#283 (text) ====
@@ -2583,6 +2583,13 @@
if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv))
{
+#ifdef PERL_COPY_ON_WRITE
+ /* While magic was saved (and off) sv_setsv may well have seen
+ this SV as a prime candidate for COW. */
+ if (SvIsCOW(sv))
+ sv_force_normal(sv);
+#endif
+
if (mgs->mgs_flags)
SvFLAGS(sv) |= mgs->mgs_flags;
else
Thread Previous
|
Thread Next