Bug #13444: compatibility breaking change: backticks is not evalueted as shell script call while in "unless" operator context

• Status: Closed
• Assignee: -
• Target version: -
• ruby -v: ruby 2.4.1p111 (2017-03-22 revision 58053) [x86_64-linux]
• Backport: 2.2: DONTNEED, 2.3: DONTNEED, 2.4: DONE

[dikderoy (Roman Bulgakov)]

Ruby versions from 2.0 to 2.3.x while executing shell script via backticks inside of "unless" construction had it executed and evalueted output as condition,
version 2.4 and onwards had condition evaluated as true but script is not executed.

I haven't find any note regarding such change in release-notes for 2.4.0 and 2.4.1

Steps to reproduce:

giving following code:

unless `touch abc.txt`
  raise 'Error'
end

Expected:

cwd has a file abc.txt created

Actual:

cwd has no file abc.txt and no exception has been raised

---

Test-case (with version comparison via rvm)

file [test.rb]:

unless `touch test-#{RUBY_VERSION}`
  raise 'Error'
end

begin
  puts $?.success?
rescue StandardError => e
  puts "rescued:#{e.message}"
end

s = `ls -l test-#{RUBY_VERSION}`
puts s.lines.length == 1 ? 'OK' : 'NOK'

exec:

$ rvm list

  rvm rubies

   ruby-2.0.0-p648 [ x86_64 ]
   ruby-2.1.10 [ x86_64 ]
   ruby-2.2.6 [ x86_64 ]
   ruby-2.3.3 [ x86_64 ]
 * ruby-2.4.0 [ x86_64 ]
=> ruby-2.4.1 [ x86_64 ]

$ rvm all --verbose do ruby test.rb

ruby-2.1.10: ruby 2.1.10p492 (2016-04-01 revision 54464) [x86_64-linux] 
true
OK

ruby-2.2.6: ruby 2.2.6p396 (2016-11-15 revision 56800) [x86_64-linux] 
true
OK


ruby-2.0.0-p648: ruby 2.0.0p648 (2015-12-16 revision 53162) [x86_64-linux] 
true
OK

ruby-2.3.3: ruby 2.3.3p222 (2016-11-21 revision 56859) [x86_64-linux] 
true
OK

ruby-2.4.0: ruby 2.4.0p0 (2016-12-24 revision 57164) [x86_64-linux] 
rescued:undefined method `success?' for nil:NilClass
NOK

ruby-2.4.1: ruby 2.4.1p111 (2017-03-22 revision 58053) [x86_64-linux] 
rescued:undefined method `success?' for nil:NilClass
NOK

$ ls -l test-*

-rw-rw-r-- 1 user user 0 апр.  17 23:50 test-2.0.0
-rw-rw-r-- 1 user user 0 апр.  17 23:50 test-2.1.10
-rw-rw-r-- 1 user user 0 апр.  17 23:50 test-2.2.6
-rw-rw-r-- 1 user user 0 апр.  17 23:50 test-2.3.3

OS: Ubuntu 14.04
cat /etc/debian_version: jessie/sid
uname -isoprvm: Linux 3.13.0-116-generic #163-Ubuntu SMP Fri Mar 31 14:13:22 UTC 2017 x86_64 x86_64 x86_64 GNU/Linux

[wanabe (_ wanabe)]

ruby does not evaluate specific kinds of expressions at if/unless cond after r54775.
NODE_XSTR is one of them.

[dikderoy (Roman Bulgakov)]

yes, You are right, it was my mistake to use it in this way, as this example is actually pointless and always True. all meaningful expression (as for ex. 1==Integer(echo 1, 10)) are evaluated correctly. Probably you should close and delete this.

[shyouhei (Shyouhei Urabe)]

Might be worth documenting.

[dikderoy (Roman Bulgakov)]

Or putting a warning, because optimization actually changes behavior.

[nobu (Nobuyoshi Nakada)]

• Status changed from Open to Closed

Applied in changeset trunk|r58398.

---

compile.c: wrong optimization

• compile.c (compile_branch_condition): expression which has side
  effects should not be eliminated.
  [ruby-core:80740] [Bug #13444]

[nobu (Nobuyoshi Nakada)]

• Backport changed from 2.2: UNKNOWN, 2.3: UNKNOWN, 2.4: UNKNOWN to 2.2: DONTNEED, 2.3: DONTNEED, 2.4: REQUIRED

[nagachika (Tomoyuki Chikanaga)]

• Backport changed from 2.2: DONTNEED, 2.3: DONTNEED, 2.4: REQUIRED to 2.2: DONTNEED, 2.3: DONTNEED, 2.4: DONE

ruby_2_4 r58634 merged revision(s) 58398.

[Eregon (Benoit Daloze)]

This is a bug, it does not execute part of the user program, even though the semantics are clearly defined.

Thanks for the fix nobu!
Maybe other case without side effects should warn, like if true ?