Bug #5865: Exception#== should return false if the classes differ

• Status: Closed
• Assignee: nobu (Nobuyoshi Nakada)
• Target version: -
• ruby -v: ruby 2.0.0dev (2011-12-31 trunk 34165) [x86_64-darwin11.2.0]

[hasari (Hiro Asari)]

Documentation says "If obj is not an Exception, returns false. Otherwise, returns true if exc and obj share same class, messages, and backtrace."

However,

$ ruby2.0 -v -e 'x=RuntimeError.new("msg"); y=ScriptError.new("msg"); p x==y'
ruby 2.0.0dev (2011-12-31 trunk 34165) [x86_64-darwin11.2.0]
true

[nobu (Nobuyoshi Nakada)]

=begin
It seems intentional.

---

r20866 | matz | 2008-12-19 08:42:00 +0900 (Fri, 19 Dec 2008) | 2 lines

• error.c (exc_equal): duck typing equal to make it transitive.
  [ruby-dev:34880]

---

But the ML ref is apparently wrong.
=end

[now (Nikolai Weibull)]

I also wrote about this in [ruby-core:41393]. An excerpt:

The commit message says that “duck typing equal to make it
transitive”, but I don’t understand what’s meant by “transitive” in
this case. I don’t think RuntimeError.new('a') should be #== to
StandardError.new('a') or vice versa.

Intentional or not at the time, this is neither expected nor desired behavior.

[hasari (Hiro Asari)]

Can this be assigned to matz for clarification?

Thanks.

[matz (Yukihiro Matsumoto)]

Hi,

In message "Re: [ruby-core:41997] [ruby-trunk - Bug #5865] Exception#== should return false if the classes differ"
on Mon, 9 Jan 2012 15:10:16 +0900, Hiro Asari [e-mail] writes:
|
|Issue #5865 has been updated by Hiro Asari.
|
|
|Can this be assigned to matz for clarification?

See [ruby-dev:34808].

						matz.

[now (Nikolai Weibull)]

On Mon, Jan 9, 2012 at 12:43, Yukihiro Matsumoto [e-mail] wrote:

In message "Re: [ruby-core:41997] [ruby-trunk - Bug #5865] Exception#== should return false if the classes differ"
   on Mon, 9 Jan 2012 15:10:16 +0900, Hiro Asari [e-mail] writes:
|
|Issue #5865 has been updated by Hiro Asari.
|
|Can this be assigned to matz for clarification?

See [ruby-dev:34808].

What’s the use case? (How often do you wrap an Exception in a SimpleDelegator?)

The fix should be to also check the #class of obj against the #class
of exc, not to simply ignore it if the rb_obj_class()es don’t match.
I’ve attached a patch.

SimpleDelegator would then need to delegate class as well, but I can’t
say whether that makes sense or not.

[nobu (Nobuyoshi Nakada)]

Hi,

(12/01/09 22:02), Nikolai Weibull wrote:

The fix should be to also check the #class of obj against the #class
of exc, not to simply ignore it if the rb_obj_class()es don’t match.
I’ve attached a patch.

Fixnum#== works fine with Delegator because of implicit conversion, so
I think Exception should also use same way.

diff --git a/error.c b/error.c
index 6844f99..7185d52 100644
--- a/error.c
+++ b/error.c
@@ -732,10 +732,14 @@ exc_equal(VALUE exc, VALUE obj)
CONST_ID(id_mesg, "mesg");

  if (rb_obj_class(exc) != rb_obj_class(obj)) {

• ID id_message, id_backtrace;

• ID id_message, id_backtrace, id_exception;
  CONST_ID(id_message, "message");
  CONST_ID(id_backtrace, "backtrace");

• CONST_ID(id_exception, "exception");

• obj = rb_check_funcall(obj, id_exception, 0, 0);

• if (obj == Qundef) return Qfalse;

• if (rb_obj_class(exc) != rb_obj_class(obj)) return Qfalse;
  mesg = rb_check_funcall(obj, id_message, 0, 0);
  if (mesg == Qundef) return Qfalse;
  backtrace = rb_check_funcall(obj, id_backtrace, 0, 0);
  diff --git a/test/ruby/test_exception.rb b/test/ruby/test_exception.rb
  index 028391d..7226fe2 100644
  --- a/test/ruby/test_exception.rb
  +++ b/test/ruby/test_exception.rb
  @@ -380,4 +380,9 @@ end.join
  load(t.path)
  end
  end

• def test_equal

• assert_equal(RuntimeError.new("a"), RuntimeError.new("a"))

• assert_not_equal(RuntimeError.new("a"), StandardError.new("a"))

• end
  end
  

--
Nobu Nakada

[matz (Yukihiro Matsumoto)]

Hi,

OK, I like this patch. Please check in.

						matz.

In message "Re: [ruby-core:42017] Re: [ruby-trunk - Bug #5865] Exception#== should return false if the classes differ"
on Tue, 10 Jan 2012 11:34:25 +0900, Nobuyoshi Nakada [e-mail] writes:
|
|Hi,
|
|(12/01/09 22:02), Nikolai Weibull wrote:
|> The fix should be to also check the #class of obj against the #class
|> of exc, not to simply ignore it if the rb_obj_class()es don$B!G(Bt match.
|> I$B!G(Bve attached a patch.
|
|Fixnum#== works fine with Delegator because of implicit conversion, so
|I think Exception should also use same way.

[now (Nikolai Weibull)]

On Tue, Jan 10, 2012 at 03:34, Nobuyoshi Nakada [e-mail] wrote:

•       obj = rb_check_funcall(obj, id_exception, 0, 0);
•       if (obj == Qundef) return Qfalse;
•       if (rb_obj_class(exc) != rb_obj_class(obj)) return Qfalse;

Ah, much better.

[nobu (Nobuyoshi Nakada)]

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

This issue was solved with changeset r34291.
Hiro, thank you for reporting this issue.
Your contribution to Ruby is greatly appreciated.
May Ruby be with you.

---

• error.c (exc_equal): try implicit conversion for delegator.
  [ruby-core:41979] [Bug #5865]

[naruse (Yui NARUSE)]

• Status changed from Closed to Assigned
• Assignee set to nobu (Nobuyoshi Nakada)

[nobu (Nobuyoshi Nakada)]

• Status changed from Assigned to Closed

This issue was solved with changeset r34299.
Hiro, thank you for reporting this issue.
Your contribution to Ruby is greatly appreciated.
May Ruby be with you.

---

• error.c (exc_equal): ignore exceptions during implicit
  conversion. [ruby-core:41979] [Bug #5865]