Rake task blows up with ‘Error: No such file - lib/**/*.rb’

[chastell]

I think this is a new bug in reek 3.2 – running reek works, running rake reek blows up with Error: No such file - lib/**/*.rb. Example from my art_decomp code:

Error: No such file - lib/**/*.rb
/home/chastell/.rubies/ruby-2.2.2/lib/ruby/2.2.0/find.rb:42:in `block in find': No such file or directory (Errno::ENOENT)
    from /home/chastell/.rubies/ruby-2.2.2/lib/ruby/2.2.0/find.rb:42:in `collect!'
    from /home/chastell/.rubies/ruby-2.2.2/lib/ruby/2.2.0/find.rb:42:in `find'
    from /home/chastell/.rubies/ruby-2.2.2/lib/ruby/2.2.0/pathname.rb:551:in `find'
    from /home/chastell/.gem/ruby/2.2.2/gems/reek-3.2/lib/reek/source/source_locator.rb:37:in `block in source_paths'
    from /home/chastell/.gem/ruby/2.2.2/gems/reek-3.2/lib/reek/source/source_locator.rb:35:in `each'
    from /home/chastell/.gem/ruby/2.2.2/gems/reek-3.2/lib/reek/source/source_locator.rb:35:in `each_with_object'
    from /home/chastell/.gem/ruby/2.2.2/gems/reek-3.2/lib/reek/source/source_locator.rb:35:in `source_paths'
    from /home/chastell/.gem/ruby/2.2.2/gems/reek-3.2/lib/reek/source/source_locator.rb:27:in `sources'
    from /home/chastell/.gem/ruby/2.2.2/gems/reek-3.2/lib/reek/cli/input.rb:39:in `sources_from_argv'
    from /home/chastell/.gem/ruby/2.2.2/gems/reek-3.2/lib/reek/cli/input.rb:18:in `sources'
    from /home/chastell/.gem/ruby/2.2.2/gems/reek-3.2/lib/reek/cli/reek_command.rb:13:in `execute'
    from /home/chastell/.gem/ruby/2.2.2/gems/reek-3.2/lib/reek/cli/application.rb:34:in `execute'
    from /home/chastell/.gem/ruby/2.2.2/gems/reek-3.2/bin/reek:12:in `<top (required)>'
    from /home/chastell/.gem/ruby/2.2.2/bin/reek:23:in `load'
    from /home/chastell/.gem/ruby/2.2.2/bin/reek:23:in `<main>'
    from /home/chastell/.gem/ruby/2.2.2/bin/ruby_executable_hooks:15:in `eval'
    from /home/chastell/.gem/ruby/2.2.2/bin/ruby_executable_hooks:15:in `<main>'

[chastell]

(I’d debug and fix outright, but I’m on vacations and can’t plan my reek time in any way, sorry.)

[troessner]

Hmm, pretty weird.

Adding a binding.pry before here: https://github.com/troessner/reek/blob/master/lib/reek/rake/task.rb#L85

gives me:

[1] pry(#<Reek::Rake::Task>)> p(*command)
"reek"
"lib/**/*.rb"
=> ["reek", "lib/**/*.rb"]

which blows up.

But if I do a

reek lib/**/*.rb

it works. Investigating...

[mvz]

I'm guessing the rake task bypasses shell globbing.

[troessner]

On the road until tomorrow - @mvz do you have time to look into it? I probably cant look into it until tomorrow evening.

[troessner]

I just did a git bisect:

git bisect start
git bisect good 6796823
git bisect bad 5596f1b

And look what it found:

( $ ((no branch, bisect started on master))) git bisect good
864f0a92fbdde7d197125d50653cb25f8a42d42c is the first bad commit
commit 864f0a92fbdde7d197125d50653cb25f8a42d42c
Author: Piotr Szotkowski <[email protected]>
Date:   Sun Jul 26 23:29:45 2015 +0200

    Access ivars only in constructors and for memoization

Checking out this commit now.

[troessner]

I considered fixing this pretty urgent since most reek integrations will be using the rake task and this will fail for everybody who upgrades or install reek fresh.

What I did:

• Found the buggy commit
• I reverted the commit in question: c5c4d07
• Made sure "rake reek" is running again and the specs are passing
• Bumped the version 3.2.1 4f962d2
• And released it

Aftermath:

Let's find out what the problem was with c5c4d07 and re-introduce it. @chastell since you're the author of this commit, would you like to take over? I quickly skimmed it but couldn't see anything obvious.

[troessner]

I think I found the bug - bugfix is up here: #644
Please review. :)

[chastell]

Great debugging! It might’ve been a tad easier if the commits were smaller – but, well… ;)

Apologies for this; as mentioned in #644 I’ll write tests for Reek::Rake::Task so that this won’t happen again.

[backus]

@troessner I think this issue should be reopened. Rake tasks that customize source_files like this one mbj/devtools: reek.rake are still broken. The current release evaluates the FileList in the initializer only while releases before 864f0a9 evaluated the FileList at runtime.

[chastell]

Oh, good point. I think using the setters is a canonical Rake task usage. I’ll fix it early next week at the latest (I’m travelling at the moment).

[troessner]

Fix up in #683 - please review.

Sorry @chastell 😆