I've taken a look into these problems. Short version: Disable HiPE (--disable-hipe) or use Linux 2.6 if you need to get Erlang 10 working properly on x86 as a matter of urgency. Note that erlang 9 has similar problems, but they don't seem to occur during startup.
The problem stems from the use of an alternate signal stack (see erts/emulator/hipe/hipe_x86_signal.c for details) interacting badly with Linuxthreads. If you look at the straces, you'll see the SIGSEGV comes immediately after SIGCHLD -- that's because it's the signal handler wrapper in Linuxthreads that's dying. If you use Linux 2.6, glibc appears to give you NPTL instead of Linuxthreads, and it works. I think the only reason Erlang 9 works is that the SIGCHLD is received before the alternate stack stuff is set up, but I'm not sure. You can see the same effect if you start up erl and hit Ctrl-C though -- it crashes rather than bringing up the break handler. If Erlang needs to handle other signals then it may well crash then too. It's probably worth asking the debian glibc folks about this problem (it'll break other programs too, this isn't a bug in erlang itself). As far as I can see, this would work if glibc is told it can assume at least a 2.4 kernel when built (see the thread around http://sources.redhat.com/ml/libc-alpha/2001-09/msg00035.html for instance), but I can't really test it and don't know if it would cause other difficulties which would prevent Debian using it. Oh, and gentoo appear to have the same problem: http://bugs.gentoo.org/show_bug.cgi?id=69041 Brian -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
