Michael Hunter wrote: > On Tue, 27 Oct 2009 09:17:15 -0700 > Renee Danson Sommerfeld <renee.sommerfeld at Sun.COM> wrote: > > >> On Tue, Oct 27, 2009 at 01:42:52PM +0000, Alan Maguire wrote: >> >>> one-liner at >>> >>> http://zhadum.east/export/ws/amaguire/nwam1-cr/webrev >>> >> Your changes certainly look right. I'm just worried that we're >> missing something, as I don't see how this code fragment could >> ever *not* cause a core dump (with umem enabled). Is it possible >> there's a different path that's usually taken? >> > > We don't set (implicitly) _IOMYBUF? > > That was my thought too. Since we always end up calling sysevent_open_channel() - where the problematic loop is - on calling sysevent_bind_handle(), the explanation for the intermittent nature of the coredump must be in the fclose() code, and given that it only free()s memory if _IOMYBUF is set for the fp, that seems the likely culprit.
Alan
