Bug #9150: Segfault in case statement execution, possibly related to refinements

[Bradley Buda]

This code consistently segfaults in 2.0.0p353:

require 'active_support/all'

case 3600
when 1.week then true
end

This is after gem install activesupport --version=3.2.13. The code works fine in 2.0.0p247.

I ran a git bisect between the two patches and I've narrowed it down to this change: https://bugs.ruby-lang.org/projects/ruby-trunk/repository/revisions/42869 . I don't know enough about ruby's internals to debug this any further, but according to LLDB the problem is a null pointer dereference in vm_eval.c:

141 {
142 VALUE ret;
143
144 if (!ci->me->def) return Qnil;
145
146 if (th->passed_block) {
147 ci->blockptr = (rb_block_t *)th->passed_block;

ci->me is null on line 144.

I can reproduce this error on both OSX and Linux. Let me know if I can provide any more info to help debug this.

[Zachary Scott]

• Status changed from Open to Assigned
• Assignee set to Tomoyuki Chikanaga

@nagachika What do you think?

[Shugo Maeda]

I've not investigated this problem yet, but Matz said implicit method calls should not honor refinements, so r42869 might be an undesirable change.

[Tomoyuki Chikanaga]

• Status changed from Assigned to Closed

duplicated. please move to #8872