אחד מ-TopicEmbed.import spec נכשל עם Ruby 3

סיכום מבוסס בינה מלאכותית

A user reports a failing spec in TopicEmbed.import with Ruby 3.0.2, where the test expects a string to include another string but fails due to differences in string freezing behavior in Ruby 3. The issue arises because interpolated strings are no longer frozen by default, even with the frozen-string-literal magic comment yahonda. The user identifies two potential fixes: using contents.dup to avoid modifying the original string or freezing the long_content string. They note that a similar fix was previously implemented but reverted, and they have opened a pull request yahonda to address the issue.

I’d like to get some help about failing TopicEmbed.import spec with Ruby 3.

Steps to reproduce

cd discourse
bundle exec rspec ./spec/models/topic_embed_spec.rb:154

Expected behavior

It should finish successfully.

Actual behavior with Ruby 3.0.2

$ bundle exec rspec ./spec/models/topic_embed_spec.rb:154
Run options: include {:locations=>{"./spec/models/topic_embed_spec.rb"=>[154]}}

Randomized with seed 41701
F

Failures:

  1) TopicEmbed.import embedded content truncation keeps everything in the imported post when truncation is disabled
     Failure/Error: expect(post.raw).to include(long_content)

       expected "<p>aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa...for the original entry at <a href=\"http://eviltrout.com/123\">http://eviltrout.com/123</a></small>" to include "<p>aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa...for the original entry at <a href='http://eviltrout.com/123'>http://eviltrout.com/123</a></small>\n"
       Diff:
       @@ -1,4 +1,7 @@
       -<p>aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa</p>\n<p>more</p>\n<hr>\n<small>This is a companion discussion topic for the original entry at <a href='http://eviltrout.com/123'>http://eviltrout.com/123</a></small>\n
       +<p>aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa</p>
       +<p>more</p>
       +<hr>
       +<small>This is a companion discussion topic for the original entry at <a href="http://eviltrout.com/123">http://eviltrout.com/123</a></small>

     # ./spec/models/topic_embed_spec.rb:158:in `block (4 levels) in <top (required)>'
     # ./spec/rails_helper.rb:279:in `block (2 levels) in <top (required)>'

Finished in 2.48 seconds (files took 2.73 seconds to load)
1 example, 1 failure

Failed examples:

rspec ./spec/models/topic_embed_spec.rb:154 # TopicEmbed.import embedded content truncation keeps everything in the imported post when truncation is disabled

Randomized with seed 41701

$

Actual behavior with Ruby 2.7.4

$ ruby -v
ruby 2.7.4p191 (2021-07-07 revision a21a3b7d23) [x86_64-linux]
$ bundle exec rspec ./spec/models/topic_embed_spec.rb:154
Run options: include {:locations=>{"./spec/models/topic_embed_spec.rb"=>[154]}}

Randomized with seed 4126
.

Finished in 2.51 seconds (files took 2.31 seconds to load)
1 example, 0 failures

Randomized with seed 4126

$

What I’ve found so far

Added some print debug code and found the long_content inside the topic_embed_spec.rb was overwritten by contents of app/models/topic_embed.rb

Print debug code

$ git diff
diff --git a/app/models/topic_embed.rb b/app/models/topic_embed.rb
index 1e9f06a322..ab407d8005 100644
--- a/app/models/topic_embed.rb
+++ b/app/models/topic_embed.rb
@@ -28,6 +28,7 @@ class TopicEmbed < ActiveRecord::Base

   # Import an article from a source (RSS/Atom/Other)
   def self.import(user, url, title, contents, category_id: nil, cook_method: nil, tags: nil)
+    p "contents.length at the begining of self.import: #{contents.length}"
     return unless url =~ /^https?\:\/\//

     if SiteSetting.embed_truncate && cook_method.nil?
@@ -102,6 +103,7 @@ class TopicEmbed < ActiveRecord::Base
       end
     end

+    p "contents.length at the end of self.import: #{contents.length}"
     post
   end

diff --git a/spec/models/topic_embed_spec.rb b/spec/models/topic_embed_spec.rb
index 8659eb7d56..1266441da7 100644
--- a/spec/models/topic_embed_spec.rb
+++ b/spec/models/topic_embed_spec.rb
@@ -153,7 +153,9 @@ describe TopicEmbed do

       it 'keeps everything in the imported post when truncation is disabled' do
         SiteSetting.embed_truncate = false
+        p "long_content.length before calling TopicEmbed.import: #{long_content.length}"
         post = TopicEmbed.import(user, url, title, long_content)
+        p "long_content.length after calling TopicEmbed.import: #{long_content.length}"

         expect(post.raw).to include(long_content)
       end
$

Result with Ruby 3.0.2

$ bundle exec rspec ./spec/models/topic_embed_spec.rb:154
Run options: include {:locations=>{"./spec/models/topic_embed_spec.rb"=>[154]}}

Randomized with seed 24988
"long_content.length before calling TopicEmbed.import: 119"
"contents.length at the begining of self.import: 119"
"contents.length at the end of self.import: 267"
"long_content.length after calling TopicEmbed.import: 267"
F

Failures:

  1) TopicEmbed.import embedded content truncation keeps everything in the imported post when truncation is disabled
     Failure/Error: expect(post.raw).to include(long_content)

       expected "<p>aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa...for the original entry at <a href=\"http://eviltrout.com/123\">http://eviltrout.com/123</a></small>" to include "<p>aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa...for the original entry at <a href='http://eviltrout.com/123'>http://eviltrout.com/123</a></small>\n"
       Diff:
       @@ -1,4 +1,7 @@
       -<p>aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa</p>\n<p>more</p>\n<hr>\n<small>This is a companion discussion topic for the original entry at <a href='http://eviltrout.com/123'>http://eviltrout.com/123</a></small>\n
       +<p>aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa</p>
       +<p>more</p>
       +<hr>
       +<small>This is a companion discussion topic for the original entry at <a href="http://eviltrout.com/123">http://eviltrout.com/123</a></small>

     # ./spec/models/topic_embed_spec.rb:160:in `block (4 levels) in <top (required)>'
     # ./spec/rails_helper.rb:279:in `block (2 levels) in <top (required)>'

Finished in 2.55 seconds (files took 3.4 seconds to load)
1 example, 1 failure

Failed examples:

rspec ./spec/models/topic_embed_spec.rb:154 # TopicEmbed.import embedded content truncation keeps everything in the imported post when truncation is disabled

Randomized with seed 24988

$

Result with Ruby 2.7.4

$ bundle exec rspec ./spec/models/topic_embed_spec.rb:154
Run options: include {:locations=>{"./spec/models/topic_embed_spec.rb"=>[154]}}

Randomized with seed 1802
"long_content.length before calling TopicEmbed.import: 119"
"contents.length at the begining of self.import: 119"
"contents.length at the end of self.import: 267"
"long_content.length after calling TopicEmbed.import: 119"
.

Finished in 2.56 seconds (files took 3.06 seconds to load)
1 example, 0 failures

Randomized with seed 1802

$

I’ve got why this spec fails only with Ruby 3.0.2.

Since Ruby 3, interpolated strings are not frozen even if frozen-string-literal magic comment is enabled.
https://bugs.ruby-lang.org/issues/17104

So here are possible ways to fix it.

  1. Use contens.dup instead of +contents
$ git diff
diff --git a/app/models/topic_embed.rb b/app/models/topic_embed.rb
index 1e9f06a322..502409d3c1 100644
--- a/app/models/topic_embed.rb
+++ b/app/models/topic_embed.rb
@@ -34,7 +34,7 @@ class TopicEmbed < ActiveRecord::Base
       contents = first_paragraph_from(contents)
     end
     contents ||= ''
-    contents = +contents << imported_from_html(url)
+    contents = contents.dup << imported_from_html(url)

     url = normalize_url(url)

$
  1. Freeze interpolated long_content
diff --git a/spec/models/topic_embed_spec.rb b/spec/models/topic_embed_spec.rb
index 8659eb7d56..a25351a586 100644
--- a/spec/models/topic_embed_spec.rb
+++ b/spec/models/topic_embed_spec.rb
@@ -142,7 +142,7 @@ describe TopicEmbed do
     describe 'embedded content truncation' do
       MAX_LENGTH_BEFORE_TRUNCATION = 100

-      let(:long_content) { "<p>#{'a' * MAX_LENGTH_BEFORE_TRUNCATION}</p>\n<p>more</p>" }
+      let(:long_content) { "<p>#{'a' * MAX_LENGTH_BEFORE_TRUNCATION}</p>\n<p>more</p>".freeze }

       it 'truncates the imported post when truncation is enabled' do
         SiteSetting.embed_truncate = true
$

I’m inclined to open a pull request to use contens.dup and found the idea was implemented but reverted via Revert "FIX: Use #dup instead of #+@ since content could be an instan… · discourse/discourse@d01c938 · GitHub

I did not find some other backgrounds for the revert, I’d like to hear from contributors/maintainers.

Opened FIX: Ruby 3 does not freeze interpolated string by yahonda · Pull Request #14567 · discourse/discourse · GitHub