Bug #21282: `FrozenError` raised by `rb_error_frozen_object` should mention real class of frozen object

• Status: Closed
• Assignee: -
• Target version: -
• ruby -v: ruby 3.5.0dev (2025-04-23T14:19:58Z fix-frozenerror-me.. 5c9dd42e27) +PRISM [x86_64-darwin24]
• Backport: 3.2: UNKNOWN, 3.3: UNKNOWN, 3.4: UNKNOWN

[sanjioh (Fabio Sangiovanni)]

Hello everyone,

I've noticed that the following snippets both raise FrozenError, but with different messages:

Object.new.freeze.instance_variable_set(:@test, true)  # can't modify frozen Object: #<Object:0x000000010b4d8aa8> (FrozenError)

Object.new.freeze.tap(&:singleton_class).instance_variable_set(:@test, true)  # can't modify frozen #<Class:#<Object:0x0000000101538a98>>: #<Object:0x0000000101538a98> (FrozenError)

As you can see, the second example mentions the singleton class of the object, while I would expect Object to be mentioned in its place, as the first example does.
From what I can see in test/ruby/test_frozen_error.rb, the expected behavior here should be to display the real class of the frozen object (as returned by Kernel#class), independently of the existence of a singleton class.

I'm working on a PR that should fix this (will post the link in the comments in a moment).

Cheers!

[sanjioh (Fabio Sangiovanni)]

PR https://github.com/ruby/ruby/pull/13161

[sanjioh (Fabio Sangiovanni)]

On a second thought, I've come to the conclusion that the benefits in clarity that a fix would provide are not worth the associated performance impact (i.e. the additional search for the real class required to format the error message).

This issue can be closed (sorry for the noise).

[byroot (Jean Boussier)]

• Status changed from Open to Closed