[perl #116819] endless loop debugging DosGlob

On Tue Jul 16 00:06:12 2013, sprout wrote:
> On Sun Feb 17 19:27:48 2013, craigberry wrote:
> > The commit in question adds an XS routine _callsite that consists
> >    entirely of:
> > 
> > SV *
> > _callsite(...)
> >     CODE:
> >        RETVAL = newSVpvn(
> >                   (char *)&cxstack[cxstack_ix].blk_sub.retop,
> >    sizeof(OP *)
> >                 );
> >     OUTPUT:
> >        RETVAL
> > 
> > I'm wondering why it's taking the address of retop since that is
> >    already of type OP*, but I may be misparsing this and don't have
> >    the time at the moment to dig deeper.
> 
> It’s reading the pointer as a string, to get the memory address of the op.
> 
> So presumably it’s getting confused because it is not stepping over the
> debugger’s call frames.  Or something like that.

Without actually confirming (or even looking at the code), I think it is
getting the address of the line in DB::sub that calls glob.

> I wonder whether Devel::Callsite has the same bug.

So yes, it probably does.

I don’t know offhand which is the best way to determine whether it
should be skipped.  I think pp_caller has something we can copy.

-- 

Father Chrysostomos


---
via perlbug:  queue: perl5 status: open
https://rt.perl.org:443/rt3/Ticket/Display.html?id=116819

• [perl #116819] endless loop debugging DosGlob by Father Chrysostomos via RT
• [perl #116819] endless loop debugging DosGlob by Father Chrysostomos via RT
• Re: [perl #116819] endless loop debugging DosGlob by Rocky Bernstein