From: Rolando Abarca <funkaster@...>
Date: 2009-08-14T01:46:45+09:00
Subject: [ruby-core:24917] Re: embedding ruby 1.9 frustration

On Thu, Aug 13, 2009 at 12:18 PM, Caleb Clausen<caleb@inforadical.net> wrote:
>
> I think this may be the right track. You've got a lot of async stuff
> happening on top of the ruby vm, and even tho you have only one thread,
> threads aren't the only possible source of reentrancy problems. Maybe you
> could try some logging to figure out what async events are happening right
> before it goes sour, and what state the vm is in when they happen...?

Ok... I think I got it:

this is what my sc_funcall used to look (wrap around rb_funcall used
with rb_protect):

VALUE sc_funcall(struct sc_funcall_param *param) {
	if (rb_obj_respond_to(param->recv, param->method_id, Qfalse))
		return rb_funcall3(param->recv, param->method_id, param->n, param->argv);
	return Qnil;
}

after removing rb_obj_respond_to I got it running smooth all the time.
Maybe rb_obj_respond_to is not "signal-safe"? It's been running for
about 20 minutes no crash so far.

-- 
Rolando Abarca
http://rolando.cl