On Sun, Apr 10, 2005 at 06:13:26PM -0400, Philip M. Gollucci wrote:
> Tim Bunce wrote:
> > If you, or someone, has something like valgrind handy (or any other
> > tool that would help with malloc/threads debugging) that would be a
> > big help.
>
> Same thing with 5.8.0, 5.8.6  both with ithreads

I'm running "perl -Mblib t/05thrclone.t" in a loop using v5.8.6 built
for i386-freebsd-thread-multi and I've had no failures yet.

> Here's the valgrind -v output... I didn't do --trace-pthreads=all ... 
> (way too much output)
> 
> I also lowered the $threads = 10 to $threads = 2 in the test script.
> I'd say it happens > 70% of the time for me.

Thanks.

> ==92812==    /home/pgollucci/dev/inst/perl/5.9.3-i/bin/perl

So this valgrind run was with a bleadperl that not released yet.
Very blead :)

> ok 6 - ... the Driver has one Kid
> ==92812== Conditional jump or move depends on uninitialised value(s)
> ==92812==    at 0x3C10F615: Perl_sv_dup (sv.c:10747)
> ==92812==    by 0x3C3FF1D8: Perl_ithread_join (threads.xs:634)
> ==92812==    by 0x3C3FFCAE: XS_threads_join (threads.xs:728)
> ==92812==    by 0x3C0F593B: Perl_pp_entersub (pp_hot.c:2816)
> ==92812== 
> ==92812== Conditional jump or move depends on uninitialised value(s)
> ==92812==    at 0x3C10F615: Perl_sv_dup (sv.c:10747)
> ==92812==    by 0x3C110907: Perl_sv_dup (sv.c:10958)
> ==92812==    by 0x3C3FF1D8: Perl_ithread_join (threads.xs:634)
> ==92812==    by 0x3C3FFCAE: XS_threads_join (threads.xs:728)

Sure looks like a perl problem. What's on & around sv.c:10747?
Is the same line of code implicated with perl 5.8.6? (possibly moved
slightly, of course.)

Tim.

Reply via email to