--auto-gen-config runs forever

[jaredbeck]

Expected behavior

--auto-gen-config should finish eventually.

Actual behavior

--auto-gen-config ran at 100% CPU for 25 minutes before I gave up.

top -pid 27995
PID    COMMAND      %CPU  TIME     #TH  #WQ  #POR MEM  PURG CMPR PGRP  PPID  STATE   BOOSTS    %CPU_ME %CPU_OTHRS UID  FAULT COW  MSGSEN MSGR SYSBSD SYSM CSW
27995  ruby         99.8  24:03.73 2/1  0    22   103M 0B   0B   27995 93817 running *0[1]     0.00000 0.00000    redacted  40598 717  13972  85   217280 1309 264006+

With --debug, the last line printed before it hangs is:

Scanning /Users/jaredbeck/redacted/lib/tasks/redacted.rake

Workaround

1. Deleted the .rake file
2. Ran --auto-gen-config
3. Restored the .rake file

Why can't --auto-gen-config handle this .rake file?

I don't know. --debug doesn't provide any clues, so I can only speculate. The file (below, redacted) has a very long string. Perhaps that's the culprit?

desc 'Redacted'
task :redacted => :environment do
  Dir.glob("#{Rails.root}/app/redacted.rb").each { |file| require file }
  redacted1 = 'redacted
300 line
28 kilobyte
string'
  # redacted 338 char long comment
  label = OpenStruct.new(redacted2: redacted1)
  redacted6 = Redacted.new(redacted3: 999).generate_pssl(redacted4: true, redacted5: label)
  ff = File.open("redacted.pdf", "w")
  ff.write(redacted6.force_encoding('utf-8'))
  ff.close
end

Steps to reproduce the problem

Try creating a file with a 28 kilobyte string?

RuboCop version

bundle exec rubocop -V
1.18.3 (using Parser 3.0.2.0, rubocop-ast 1.8.0, running on ruby 2.6.3 x86_64-darwin20)

[jonas054]

I have not been able to reproduce. @jaredbeck Do you also get it if there's no .rubocop.yml file? I'm thinking that it's perhaps related to your configuration somehow.

[jaredbeck]

Hi Jonas, Please try the following minimal reproduction:

mkdir issue_9946 && cd issue_9946
ruby -v # 2.7.2
rubocop -v # 1.18.3
curl -O https://gist.githubusercontent.com/jaredbeck/cb1b6e1b826d03a30184e539a84bbae4/raw/9103bae827a4756510e190227fa48e788c49bffe/rubocop_issue_9946.rb
rm .rubocop*
rubocop --auto-gen-config --debug

It should print

...
Phase 2 of 2: run all cops
...
Scanning .../rubocop_issue_9946.rb

And then run forever, at 100% CPU.

Critical problematic input

A long comment (in this case, 13228 chars), of random content.

Suggestions

Add a Timeout? If --auto-gen-config times out on a specific file, add it to AllCops: Exclude:?

[dvandersluis]

Do you have the same behaviour if you just run rubocop (without --auto-gen-config) on that file? My guess is that one of the cops is not completing due to the super long string (maybe LineLength?)

[jaredbeck]

Do you have the same behaviour if you just run rubocop (without --auto-gen-config) on that file?

Yes.

rm .rubocop*
rubocop --debug # 1.18.4
For /.../redacted: Default configuration from /redacted/gems/rubocop-1.18.4/config/default.yml
Inspecting 1 file
Scanning /.../redacted/redacted.rb

Also, the ruby version doesn't seem to matter. I reproduced this with both ruby 2.7.2 and 3.0.2.

[jonas054]

I've been able to reproduce the problem. Thanks @jaredbeck! And I've found that it's the Style/CommentedKeyword cop that hangs. It's actually possible that it will finish at some point, but the time is spent in regexp matching, trying to match /(?<keyword>\S+).*#/ against the very long comment in this example.

It's a regular expression that can't always be matched in reasonable time. I suggest we correct this cop rather than monitoring time-outs in cops.

[jonas054]

I'll try to fix it.

[jaredbeck]

Thanks for the quick fix, everyone.