ruby-core@ruby-lang.org archive (unofficial mirror)
 help / color / mirror / Atom feed
From: mkaeppler@gitlab.com
To: ruby-core@ruby-lang.org
Subject: [ruby-core:104788] [Ruby master Bug#17806] Bad interaction between method cache, prepend, and refinements
Date: Thu, 05 Aug 2021 11:49:30 +0000 (UTC)	[thread overview]
Message-ID: <redmine.journal-93123.20210805114929.16806@ruby-lang.org> (raw)
In-Reply-To: redmine.issue-17806.20210416032511.16806@ruby-lang.org

Issue #17806 has been updated by mk (Matthias Käppler).


Hi @alanwu,

at GitLab we are currently working on making our application Ruby 3 ready.

We are running into a strange issue in our test suite where refinements fails to be applied when everything suggests they should be.

During my research I stumbled on this issue here and I was wondering if you could give me your thoughts on whether this could be a Ruby VM bug.

I summarized in this issue what the problem is: https://gitlab.com/gitlab-org/gitlab/-/issues/337614

To recap the high-level highlights and observations:

- We use rspec-parameterized and its table-syntax extension (which overrides the pipe `|` operator using refinements)
- Whenever we run a test that composes a test table using integer arguments, Ruby will invoke the built-in bitwise-OR instead of the refinement, thus breaking the test
- When I add another refinement to Integer at the top of the test file, even when not `using` it, the test will start to pass

This leads me to believe it could be due to Ruby's method table being messed up somehow, since it is very suspicious that by refining a class with an unused method, another refinement should suddenly start to apply. Since this changeset here touches the caching behavior, I thought you might have an idea whether it could be related.

We also verified that this happens not just with 3.0.2, but also 3.0.1, but works on 2.7, so it is likely not strictly related to this particular change. But since I have no evidence that it is in fact a Ruby bug, I was also hesitant to open a new issue.

I have so far not been able to produce a minimal executable test case either.

Any hints and thoughts for how to debug this would be appreciated!

----------------------------------------
Bug #17806: Bad interaction between method cache, prepend, and refinements
https://bugs.ruby-lang.org/issues/17806#change-93123

* Author: alanwu (Alan Wu)
* Status: Closed
* Priority: Normal
* ruby -v: ruby 3.0.1p64 (2021-04-05 revision 0fb782ee38) [x86_64-darwin19]
* Backport: 2.6: UNKNOWN, 2.7: UNKNOWN, 3.0: DONE
----------------------------------------
I'm running into a couple of issues with Ruby 3's new method cache and
refinements.

The first script raises `SystemStackError` unexpectedly:

```ruby
module R1
  refine Hash do
    def foo; :r1; end
  end
end
class Hash
  prepend(Module.new)
end
class Hash
  def foo; end
end
{}.method(:foo) # put it on pCMC
module R2
  refine Hash do
    def foo; :r2; end
  end
end
{}.foo # SystemStackError
```

The second script calls the wrong method:

```ruby
klass = Class.new { def foo; end }
_refinement = Module.new do
  refine(klass) { def foo; :refined; end }
end
klass.prepend(Module.new)
klass.new.foo # cache foo
klass.define_method(:foo) { :second }
p klass.new.foo # prints nil. False caching.
```

I submitted a GitHub PR to fix the issue: https://github.com/ruby/ruby/pull/4386



-- 
https://bugs.ruby-lang.org/

Unsubscribe: <mailto:ruby-core-request@ruby-lang.org?subject=unsubscribe>
<http://lists.ruby-lang.org/cgi-bin/mailman/options/ruby-core>

  parent reply	other threads:[~2021-08-05 11:49 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-04-16  3:25 [ruby-core:103469] [Ruby master Bug#17806] Bad interaction between method cache, prepend, and refinements XrXr
2021-05-29  5:57 ` [ruby-core:104096] " nagachika00
2021-08-05 11:49 ` mkaeppler [this message]
2021-08-05 15:29 ` [ruby-core:104791] " XrXr
2021-08-12 13:31 ` [ruby-core:104897] " mkaeppler

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-list from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

  List information: https://www.ruby-lang.org/en/community/mailing-lists/

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=redmine.journal-93123.20210805114929.16806@ruby-lang.org \
    --to=ruby-core@ruby-lang.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).