Segfault when trying to print last element of nqp::backtrace()

[usev6]

The Problem

While hunting a NullPointerException on the JVM backend I've stumbled upon a segfault on MoarVM. The nqp::backtrace() that is used to produce the segfault stems from src/core.c/CallFrame.pm6 -- it's actually used in sub callframe.

$ ./rakudo-m -e 'use nqp; my $bt := nqp::backtrace(nqp::handle(nqp::die(""),"CATCH",nqp::exception)); say nqp::elems($bt)'
11
$ ./rakudo-m -e 'use nqp; my $bt := nqp::backtrace(nqp::handle(nqp::die(""),"CATCH",nqp::exception)); say $bt[10]'
Segmentation fault (core dumped)

Looking a bit further, it turns out that the key sub in that last element of that backtrace contains a NULL:

$ ./rakudo-m -e 'use nqp; my $bt := nqp::backtrace(nqp::handle(nqp::die(""),"CATCH",nqp::exception)); my $sub := nqp::atkey(nqp::getattr($bt[10],Map,"\$!storage"),"sub"); dd nqp::isnull($sub)'
1

If I'm not mistaken this NULL comes from this line: https://github.com/MoarVM/MoarVM/blob/c7e75be60a/src/core/exceptions.c#L576. Doing a NULL check there and falling back to tc->instance->VMNull seems to avoid the segmentation fault. But I don't know if that would be the correct fix.

Expected Behavior

Obviously, there should be no segfault. I'm not sure what the correct value for sub should be at that point.

Compiler version (perl6 -v):

This is Rakudo version 2020.02.1-289-g2cbf58372 built on MoarVM version 2020.02.1-71-g1bd5f19e0
implementing Raku 6.d.

[jnthn]

Doing a NULL check there and falling back to tc->instance->VMNull seems to avoid the segmentation fault. But I don't know if that would be the correct fix.

yes, it would.