Andreas Jaeger <aj(a)arthur.rhein-neckar.de&gt; writes in xemacs-beta(a)xemacs.org: ... I was not in the few moments I had before the machine died, but the comment about MALLOC_CHECK_ rings a bell. I do remember getting crashes that way when I was testing. The voodoo_free_hook method had the property that full malloc checking worked without crashing. O.K. I see no reason to change how we do things at this time. We should be immune to the reported crash and if a better method of curing it becomes available some time in the future, we can switch to it. -- I protest the NATO war in Yugoslavia.