Re: Segfault in extend.c/ parrot_pmc_get_pointer

On Monday, December 15, 2003, at 12:58  pm, Leopold Toetsch wrote:

> Arthur Bergman <[email protected]> wrote:
>
>> #0  0x081cef45 in Parrot_PMC_get_pointer (interp=0x82d7f78,
>> pmc=0x83333a0)
>
> That looks like the vtable could be corrupted.
> $ p *pmc
> $ p *pmc->vtable
>


So, after a couple of days struggling to find a lost off64_t I am able 
to try again, (plus side is, I now build a threaded perl to make it 
work (otherwise nasty IO problems hit me)).

So the result.

Most likely,

(gdb) p *pmc
$2 = {obj = {u = {b = {bufstart = 0x4212dfd8, buflen = 137797904}, ptrs 
= {_struct_val = 0x4212dfd8, _pmc_val = 0x836a110},
       int_val = 1108533208, num_val = 4.2834352799906154e-269, 
string_val = 0x4212dfd8}, flags = 1886221358, pobj_version = 0}, vtable 
= 0x18,
   pmc_ext = 0x10}

Of course followed by a

(gdb) p *pmc->vtable
Cannot access memory at address 0x18


If it wasn't for pobj_version = 0 I would suggest it was all just 
random memory I was looking at.

>> Any suggestions?
>
> Maybe the PMC got destroyed by a DOD run. You can check that by
> disabling DOD after Parrot_new(). (Parrot has a commandline option -G
> for this). Also the pmc's flags should reflect that.
>

I doubt this happens since I never give parrot a chance to do anything, 
there is no step that I call into parrot, just

     PL_Parrot = Parrot_new(0);
     Parrot_init(PL_Parrot);

then a bit later


    Parrot_Int  type = Parrot_PMC_typenum(PL_Parrot, "Perl5LV");
    Parrot_PMC_get_pointer(PL_Parrot, SvANY(sv));


Arthur