|
From: Julian S. <js...@ac...> - 2006-11-01 19:18:42
|
> ==26931== Process terminating with default action of signal 6 (SIGABRT): > dumping core > ==26931== at 0x4159805: raise (in /lib/tls/libc-2.3.3.so) > ==26931== by 0x415B1E8: abort (in /lib/tls/libc-2.3.3.so) > ==26931== by 0x40CD7C4: __cxxabiv1::__terminate(void (*)()) > (in /usr/lib/libstdc++.so.5.0.6) > ==26931== by 0x40CD801: std::terminate() > (in /usr/lib/libstdc++.so.5.0.6) > ==26931== by 0x40CD981: __cxa_throw (in > /usr/lib/libstdc++.so.5.0.6) > ==26931== by 0x40CDBC2: operator new(unsigned) > (in /usr/lib/libstdc++.so.5.0.6) > ==26931== by 0x817BA03: boost::any::any<float>(float const&) (any.hpp:33) > ==26931== by 0x817B9D9: Numerica::PropertyWrapper ... > ==26931== by 0x817B968: void Numerica::Properties::setProperty ... > ==26931== by 0x81797A4: void Numerica::ExternalProperties ... > ==26931== by 0x81782D6: Numerica::ONIRMessageTranslator: > ==26931== by 0x8176FA9: Numerica::ONIRMessageTranslator ... My first reaction to this is that operator new(unsigned) decided to throw an exception and terminate the program with SIGABRT. Why? I don't know, but one obvious guess is that it thought it was out of memory for whatever reason. We had an obscure runs-on-memcheck fails-on-other-tools bug in 3.2.0 but that was fixed in 3.2.1. Maybe this is a new variant. What happens if you run it with --tool=none (instead of --tool=callgrind) ? Does it work? If no, try these combinations: --tool=none --vex-iropt-level=1 --tool=none --vex-iropt-level=0 --tool=none --sanity-level=3 --tool=none --sanity-level=4 J |