Re: Bug in ORBit_free_via_TypeCode?

[Michael Meeks <michael ximian com>]

Hi Richard,

On 27 Jun 2001, Richard Hult wrote:
> after trying to debug a weird crash in the bonobo listener skels, when
> passing a sequence of integers, I finally found what seems to be a bug
> in ORBit.

	Can you post the stack trace for the crash ? [ possibly export
_MALLOC_CHECK=2 and linking vs. -lefence might give a better trace ].

	Or better still, if you can get it down to a simple test case,
then we can not only fix it, but ensure the test runs for every release in
future.

> Take a look at ORBit_free_via_TypeCode (), allocators.c:220
> 
> case CORBA_tk_sequence:
> 	{
> 		CORBA_sequence_octet *pval = mem;
> 		if(pval->_release)
> 			CORBA_free(pval->_buffer);
> 
> 		retval = (guchar *)mem + sizeof(CORBA_sequence_octet);
> 	}
> 
> It seems like a sequence is always treated like a sequence of octets,
> but in my case the sequence contained 32-bit integers. I'm not very
> familiar with this particular parts of ORBit, so unfortenately, I
> don't have a patch.

	the sequence_octet is a base type that is the same for all
sequences, the actual buffer in pval->_buffer is a pointer, so
independant of type all sequences share the same footprint.

	Regards,

		Michael.

-- 
 mmeeks@gnu.org  <><, Pseudo Engineer, itinerant idiot