Bug #4473: Calling return within begin still executes else

• Status: Closed
• Assignee: mame (Yusuke Endoh)
• Target version: 1.9.3
• ruby -v: -

[thuss (Todd Huss)]

=begin
I see this issue in 1.9.2-p0 through the current version 1.9.2-p136 on Mac OS X. The following code prints 'else executed'. Whereas in Ruby 1.8.x it would not execute the else statement if you called return from within a begin block.

In summary, I would expect the following code to not print anything, but in Ruby 1.9.2 it actually prints 'else executed':

This code prints 'else executed'

def test_return_in_method
begin
return
rescue
puts 'rescue executed'
else
puts 'else executed'
end
end
test_return_in_method
=end

[thuss (Todd Huss)]

• File ruby-1.9.2-return-else.rb ruby-1.9.2-return-else.rb added

=begin
The code didn't paste so well so I'm attaching the source code
=end

[mayank.kohaley (Mayank Kohaley)]

=begin
I can see the same issue happening on Windows also.
=end

[ko1 (Koichi Sasada)]

• ruby -v changed from ruby 1.9.2p136 (2010-12-25 revision 30365) [x86_64-darwin10.6.0] to -

Hi,

I recognize this issue. Matz, what do you think about it?

(2011/04/06 21:12), Mayank Kohaley wrote:

Issue #4473 has been updated by Mayank Kohaley.

I can see the same issue happening on Windows also.

Bug #4473: Calling return within begin still executes else
http://redmine.ruby-lang.org/issues/4473

Author: Todd Huss
Status: Open
Priority: Normal
Assignee:
Category:
Target version:
ruby -v: ruby 1.9.2p136 (2010-12-25 revision 30365) [x86_64-darwin10.6.0]

I see this issue in 1.9.2-p0 through the current version 1.9.2-p136 on Mac OS X. The following code prints 'else executed'. Whereas in Ruby 1.8.x it would not execute the else statement if you called return from within a begin block.

In summary, I would expect the following code to not print anything, but in Ruby 1.9.2 it actually prints 'else executed':

This code prints 'else executed'

def test_return_in_method
begin
return
rescue
puts 'rescue executed'
else
puts 'else executed'
end
end
test_return_in_method

--
// SASADA Koichi at atdot dot net

[matz (Yukihiro Matsumoto)]

Hi,

In message "Re: [ruby-core:36926] Re: [Ruby 1.9 - Bug #4473] Calling return within begin still executes else"
on Fri, 10 Jun 2011 22:36:19 +0900, SASADA Koichi [e-mail] writes:

|I recognize this issue. Matz, what do you think about it?

I think it should print nothing, like 1.8 does.

						matz.

[ko1 (Koichi Sasada)]

• Assignee set to ko1 (Koichi Sasada)

Yugui-san,

This fix will introduce incompatibility from 1.9.2. Is it acceptable?

[ko1 (Koichi Sasada)]

(2011/06/10 22:56), Yukihiro Matsumoto wrote:

In message "Re: [ruby-core:36926] Re: [Ruby 1.9 - Bug #4473] Calling return within begin still executes else"
on Fri, 10 Jun 2011 22:36:19 +0900, SASADA Koichi [e-mail] writes:

|I recognize this issue. Matz, what do you think about it?

I think it should print nothing, like 1.8 does.

OK. I'll check it.

--
// SASADA Koichi at atdot dot net

[nobu (Nobuyoshi Nakada)]

I bet that it is a bug in 1.8.

[ko1 (Koichi Sasada)]

(2011/06/11 16:21), Nobuyoshi Nakada wrote:

I bet that it is a bug in 1.8.

Matz, could you give me a comment, again?

--
// SASADA Koichi at atdot dot net

[matz (Yukihiro Matsumoto)]

Hi,

In message "Re: [ruby-core:37092] Re: [Ruby 1.9 - Bug #4473] Calling return within begin still executes else"
on Mon, 13 Jun 2011 23:04:46 +0900, SASADA Koichi [e-mail] writes:

|(2011/06/11 16:21), Nobuyoshi Nakada wrote:
|> I bet that it is a bug in 1.8.
|
|Matz, could you give me a comment, again?

He didn't disclosed the reason why it was a 1.8 bug. If he has
a reasonable reason, I would accept it.

						matz.

[nahi (Hiroshi Nakamura)]

• Target version set to 1.9.3

[mame (Yusuke Endoh)]

• Assignee changed from ko1 (Koichi Sasada) to mame (Yusuke Endoh)

Hello,

|(2011/06/11 16:21), Nobuyoshi Nakada wrote:
|> I bet that it is a bug in 1.8.
|
|Matz, could you give me a comment, again?

He didn't disclosed the reason why it was a 1.8 bug. If he has
a reasonable reason, I would accept it.

I have no idea which behavior is better. But, at least, the
behavior of 1.9 seems to be caused by a wrong optimization:

def foo
begin
return
rescue
else
p :foo!
end
end

foo #=> :foo!

def foo
begin
return
rescue
else
p :foo!
end
p :BOO # dummy sentence
end

foo #=> nothing output, as OP expected

I'll commit the following patch unless there is objection.
Let's discuss the better spec in another thread, if needed.

diff --git a/parse.y b/parse.y
index 9769bb9..ffa649b 100644
--- a/parse.y
+++ b/parse.y
@@ -8713,6 +8713,10 @@ reduce_nodes_gen(struct parser_params *parser, NODE **body)
if (!subnodes(nd_head, nd_resq)) goto end;
break;
case NODE_RESCUE:

•   if (node->nd_else) {
  

•   body = &node->nd_resq;
  

•   break;
  

•   }
    if (!subnodes(nd_head, nd_resq)) goto end;
    break;
  default:
  

--
Yusuke Endoh [e-mail]

[mame (Yusuke Endoh)]

• Status changed from Open to Closed
• % Done changed from 0 to 100

This issue was solved with changeset r32483.
Todd, thank you for reporting this issue.
Your contribution to Ruby is greatly appreciated.
May Ruby be with you.

---

• parse.y (reduce_nodes_gen): NODE_RETURN in rescue body must not be
  reduced when there is an else clause. This caused bizarre behavior
  in [Bug #4473] [ruby-core:35629] [ruby-core:37884].