Bug: segfault when using super and method_missing

ruby-talk
1

Hello,

The following code produces a segfault with ruby-1.8.4 from gentoo, as well as with ruby-1.8 from cvs and 1.9 from cvs. There appears to be an issue when super is called in the subclass and the parent class has method_missing defined. Strangely, if super is called with explicit arguments, no segfault occurs. Likewise, if the method definition in the subclass is modified slightly (see below) the segfault is avoided.

Thanks,
Brad

···

------------------------------------
class BaseClass
   def method_missing(*args)
     p args
   end
end

class Article < BaseClass

   # if this is defined as title=(arg) the segfault does not occur
  def title=(*args)
     super(args) # works
     super(*args) # works
     super # segfault...
   end
end

a = Article.new
a.body = 'body'
a.title = 'foo'
-------------------------------

2

Hi,

This works for me. I've tried both:

ruby 1.8.2 (2004-12-25) [i386-mswin32] (Windows)
ruby 1.8.4 (2005-12-24) [i386-linux] (gentoo linux)

Just a guess, but maybe you compiled ruby with to much optimizations in
CFLAGS ? Mine are

CFLAGS="-O2 -fomit-frame-pointer"

Hope this helps...

Guillaume

···

Brad Hilton wrote:

Hello,

The following code produces a segfault with ruby-1.8.4 from gentoo,
as well as with ruby-1.8 from cvs and 1.9 from cvs. There appears to
be an issue when super is called in the subclass and the parent class
has method_missing defined. Strangely, if super is called with
explicit arguments, no segfault occurs. Likewise, if the method
definition in the subclass is modified slightly (see below) the
segfault is avoided.

Thanks,
Brad

------------------------------------
class BaseClass
   def method_missing(*args)
     p args
   end
end

class Article < BaseClass

   # if this is defined as title=(arg) the segfault does not occur
  def title=(*args)
     super(args) # works
     super(*args) # works
     super # segfault...
   end
end

a = Article.new
a.body = 'body'
a.title = 'foo'
-------------------------------

3

Hi,

The following code produces a segfault with ruby-1.8.4 from gentoo,
as well as with ruby-1.8 from cvs and 1.9 from cvs. There appears to
be an issue when super is called in the subclass and the parent class
has method_missing defined. Strangely, if super is called with
explicit arguments, no segfault occurs. Likewise, if the method
definition in the subclass is modified slightly (see below) the
segfault is avoided.

A bug was in super without any argument. The patch attached should
fix the bug. Thank you for reporting it.

              matz.

--- eval.c 3 Mar 2006 17:39:26 -0000 1.616.2.165
+++ eval.c 23 Mar 2006 01:48:22 -0000
@@ -5578,7 +5578,18 @@ method_missing(obj, id, argc, argv, call
     }
+ if (argc < 0) {
+ VALUE tmp;

- nargv = ALLOCA_N(VALUE, argc+1);
- nargv[0] = ID2SYM(id);
- MEMCPY(nargv+1, argv, VALUE, argc);
+ argc = -argc-1;
+ tmp = splat_value(argv[argc]);
+ nargv = ALLOCA_N(VALUE, argc + RARRAY(tmp)->len + 1);
+ MEMCPY(nargv+1, argv, VALUE, argc);
+ MEMCPY(nargv+1+argc, RARRAY(tmp)->ptr, VALUE, RARRAY(tmp)->len);
+ argc += RARRAY(tmp)->len;

+ }
+ else {
+ nargv = ALLOCA_N(VALUE, argc+1);
+ MEMCPY(nargv+1, argv, VALUE, argc);
+ }
+ nargv[0] = ID2SYM(id);
     return rb_funcall2(obj, missing, argc+1, nargv);

···

In message "Re: bug: segfault when using super and method_missing" on Thu, 23 Mar 2006 10:09:37 +0900, Brad Hilton <bhilton@vpop.net> writes:

4

Brad Hilton wrote:

Hello,

The following code produces a segfault with ruby-1.8.4 from gentoo, as

It's ok on

ruby 1.8.4 (2005-12-24) [i686-linux]

built from source with the default options (on ubuntu).

···

--
      vjoel : Joel VanderWerf : path berkeley edu : 510 665 3407

5

Hi,

···

In message "Re: bug: segfault when using super and method_missing" on Thu, 23 Mar 2006 10:50:01 +0900, Joel VanderWerf <vjoel@path.berkeley.edu> writes:

It's ok on

ruby 1.8.4 (2005-12-24) [i686-linux]

built from source with the default options (on ubuntu).

It happens only on CVS top.

              matz.

6

Thanks for your help and the quick patch!

-Brad

···

On Mar 22, 2006, at 5:49 PM, Yukihiro Matsumoto wrote:

A bug was in super without any argument. The patch attached should
fix the bug. Thank you for reporting it.