[svn:parrot] r36739 - branches/rename_pccinvoke/src/call

Andrew Whitworth wknight8111 at gmail.com
Mon Feb 16 12:44:26 UTC 2009


It may not need to be registered, no. However, it was being
unregistered at one point so there was a weird lack of symmetry in the
first place. Plus, I'm still seeing some strange issues where
Parrot_pcc_invoke_method_from_c_args and GC runs mix, so I want to
make extra-certain that everything involved in the call is nailed down
tight. This doesn't actually fix anything right now, but it does rule
out one potential problem.

--Andrew Whitworth

On Sun, Feb 15, 2009 at 5:07 PM, chromatic <chromatic at wgz.org> wrote:
> On Saturday 14 February 2009 17:01:55 whiteknight at svn.parrot.org wrote:
>
>> Log:
>> [rename_pccinvoke] registering the sigobject. It was being unregistered
>> later, but never registered. I added this to double-check that the GC knew
>> it existed.
>> Modified: branches/rename_pccinvoke/src/call/pcc.c
>
>> ===========================================================================
>> --- branches/rename_pccinvoke/src/call/pcc.c  Sun Feb 15 00:00:32
>> 2009  (r36738) +++ branches/rename_pccinvoke/src/call/pcc.c   Sun Feb 15
>> 01:01:54 2009 (r36739) @@ -374,6 +374,7 @@
>>      INTVAL       in_return_sig = 0;
>>      INTVAL       i;
>>
>> +    gc_register_pmc(interp, call_object);
>
> It shouldn't need registering or unregistering at all, at least if it's
> reachable from the root set or visible somewhere within the stack chunks that
> Parrot scans.
>
> -- c
> _______________________________________________
> http://lists.parrot.org/mailman/listinfo/parrot-dev
>


More information about the parrot-dev mailing list