github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm
Set by AlexDaniel on 12 June 2018.
moon-child odd, enabling asan has no effect 01:36
timotimo does the whole rakudo test suite (not necessarily the spec tests) pass? 02:39
moon-child yes 03:10
moon-child now I'm confused 04:47
I got it running with debugsyms, and the last bit of code in moar before it's libc error detection is nativecall_dyncall.c:362
which is break;
ah-hmmm. I recompiled just that file without optimizations, to see if that would change the location of the error. With that change, there's no stack smashing error 04:50
here we go, ubsan to the rescue! 04:53
runtime error: member access within misaligned address ... for type MVMObject/MVMCollectable/MVMCode/MVMCodeBody/MVMStaticFrame */MVMSTable * 04:54
lines 86, 91, and 207 of nativecall_dyncall.c 04:55
nine moon-child: that error may indicate that your native function returns NULL instead of a function pointer 08:37
moon-child: apparently we don't have any NULL checks there 08:38
moon-child nine: I don't think so. Execution enters the function properly and I can do stuff inside of it. It's just the return that causes those errors. And ubsan mentions addresses of actual pointers, and it all works fine without optimizations 20:34

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!