Re: [tabled patch] abstract out TCP-write code

On 09/23/2010 09:57 AM, Pete Zaitcev wrote:
> Side effect or not, if one applies your patch and executes
> "export MALLOC_PERTURB_=43" command before "make check",
> the result is a crash:

Yes, it is clear my patch re-introduces the problem.  I thanked you for 
pointing that out in the original reply.  :)

The point was, that does not invalidate the approach of storing 
write_compl_q in struct atcp_wr_state.

It merely highlights that a further fix (to the FSM? refcount client? 
etc.) is required.

	Jeff