develooper Front page | perl.perl6.internals | Postings from May 2008

Re: [perl #53714] [PATCH] parrot segfaults when dyna-loading main program

From:
chromatic
Date:
May 5, 2008 22:43
Subject:
Re: [perl #53714] [PATCH] parrot segfaults when dyna-loading main program
On Monday 05 May 2008 13:50:51 NotFound wrote:

> On Mon, May 5, 2008 at 2:52 AM, via RT Tom Erdevig
>
> <parrotbug-followup@parrotcode.org> wrote:
> >  Parrot segfaults running this:
> >  .sub _ :main
> >     null S0
> >     loadlib P0, S0
> >  .end
> >  The segfault is caused by src/dynext.c:Parrot_load_lib setting
> >  its local lib_name variable to NULL and then passing it along
> >  to run_init_lib, and thence to store_lib_pmc, who tries to set
> >  the new library PMC's '_lib_name' property using this NULL,
> >  leading to the crash.
> >
> >  The attached patch fixes this in Parrot_load_lib by setting lib_name
> >  to the empty string instead of NULL when there is no library name.
>
> I tried another way. The segfault is at
> pmc/string.pmc:set_string_native. If I insert an assertion od
> non-nullness of the 'value' parameter of this function parrot does not
> build, then I suppose the function must accept a NULL value. All
> intermediate functions have also his corresponding parameter not
> declared as nonnull, so I assume that the NULL argument must be
> allowed.
>
> The failed line is:
>
> if (PObj_constant_TEST(SELF) && !PObj_constant_TEST(value)) {
>
> If I change it to:
>
> if (PObj_constant_TEST(SELF) && (value && !PObj_constant_TEST(value))) {
>
> The segfault disappear and all tests pass.
>
> The attached patch contains this change.

I prefer this patch.  It fixes the problem closer to its source.  Does it work 
for you?

-- c





Comments to Ask Bjørn Hansen at ask@perl.org | Group listing | About