Odd break behavior?

ruby-talk
1

Run the code snippet below:

class A
  def outer
    inner do
      break
    end
  end

  def inner
    while true
      yield
    end
    puts "Should never appear"
  end
end

['something'].each do
  A.new.outer
  puts "1"
end

A.new.outer
puts "2"

Then comment out the something each block and rerun it. The first
run will look like:
1
2

The second run will look like:
Should never appear
2

For consistencies sake, I would have expected the first run to look like:
1
Should never appear
2

By virtue of previously calling outer in a block the second call to outer
now has different behavior (even though to me they look totally unrelated).
Can someone explain why this is?

-Tom

···

--
+ http://www.tc.umn.edu/~enebo +---- mailto:enebo@acm.org ----+

Thomas E Enebo, Protagonist | "Luck favors the prepared |
                             > mind." -Louis Pasteur |

2

Hi,

···

In message "Re: Odd break behavior?" on Sun, 19 Feb 2006 01:40:48 +0900, Thomas E Enebo <enebo@acm.org> writes:

Run the code snippet below:

<snip>

It seems to be a bug. Thank you for finding it. It's too complex to
fix it in a minute.

              matz.

3

On Sun, 19 Feb 2006, Yukihiro Matsumoto defenestrated me:

···

In message "Re: Odd break behavior?" > on Sun, 19 Feb 2006 01:40:48 +0900, Thomas E Enebo <enebo@acm.org> writes:

> Run the code snippet below:

<snip>

It seems to be a bug. Thank you for finding it. It's too complex to
fix it in a minute.

  I am trying to determine the proper behavior to fix break in jruby.
Was what I expected the proper behavior?

-Tom

--
+ http://www.tc.umn.edu/~enebo +---- mailto:enebo@acm.org ----+

Thomas E Enebo, Protagonist | "Luck favors the prepared |
                             > mind." -Louis Pasteur |

4

It seems to be a bug. Thank you for finding it. It's too complex to
fix it in a minute.

moulon% diff -u eval.c~ eval.c
--- eval.c~ 2005-12-20 14:41:47.000000000 +0100
+++ eval.c 2006-02-18 18:46:29.000000000 +0100
@@ -752,7 +752,7 @@
#define BLOCK_LAMBDA 2

static struct BLOCK *ruby_block;
-static unsigned long block_unique = 0;
+static unsigned long block_unique = 1;

#define PUSH_BLOCK(v,b) do { \
     struct BLOCK _block; \
moulon%

no ?

Guy Decoux

5

Hi,

···

In message "Re: Odd break behavior?" on Sun, 19 Feb 2006 03:02:04 +0900, ts <decoux@moulon.inra.fr> writes:

-static unsigned long block_unique = 0;
+static unsigned long block_unique = 1;

no ?

Yes! You must be a genius. Thank you.

              matz.

6

Hi,

···

In message "Re: Odd break behavior?" on Sun, 19 Feb 2006 02:14:37 +0900, Thomas E Enebo <enebo@acm.org> writes:

I am trying to determine the proper behavior to fix break in jruby.
Was what I expected the proper behavior?

"break" should always terminate the lexically innermost loop or block.
In this case, invocation of "inner" should have been terminated. A
bug cause termination of while loop (the closest loop in dynamic call
graph) in the definition of "inner" erroneously.

              matz.