On Tuesday 18 November 2008 04:27:38 James Keenan via RT wrote: > On Mon Nov 17 19:30:10 2008, [EMAIL PROTECTED] wrote:
> > I need to see more, but I think I know what this is. Use the 'bt' > > command in gdb to show the whole backtrace. Hopefully it includes > > Parrot_really_destroy(). > Sorry to have forgotten that step. See attachment. A real fix is getting order-of-destruction working, but I turned the concurrency scheduler into a constant PMC and added a live flag flip right before marking the concurrency scheduler, so that its custom mark always runs so that its kids always get marked as live appropriately. Fixed in r32841. -- c