Bug #13005: Inline rescue is inconsistent when rescuing NoMethodError

• Status: Closed
• Assignee: matz (Yukihiro Matsumoto)
• Target version: -
• ruby -v: ruby 2.4.0dev (2016-11-16 trunk 56815) [x86_64-darwin15]
• Backport: 2.1: UNKNOWN, 2.2: UNKNOWN, 2.3: UNKNOWN

[josh.cheek (Josh Cheek)]

o = "some object"

class << o
  private
  attr_accessor :a
end

# GETTER can be rescued in a begin-block or inline
begin   o.a     
rescue; $!.class        # => NoMethodError
end
o.a rescue $!.class     # => NoMethodError

# SETTER can be rescued in a begin-block but NOT inline!
begin   o.a = 1
rescue; $!.class        # => NoMethodError
end
o.a = 1 rescue $!.class

# ~> NoMethodError
# ~> private method `a=' called for "some object":String
# ~> Did you mean?  a
# ~>
# ~> program.rb:18:in `<main>'

[matz (Yukihiro Matsumoto)]

• Status changed from Open to Closed

o.a = 1 rescue $!.class

is parsed as

o.a = (1 rescue $!.class)

Try

(o.a = 1) rescue $1.class

Matz.

[duerst (Martin Dürst)]

• Status changed from Closed to Open
• Assignee set to matz (Yukihiro Matsumoto)

@matz (Yukihiro Matsumoto): Is there a good reason that o.a = 1 rescue $!.class is interpreted as o.a = (1 rescue $1.class), while it looks to me as if o.a = 1 if $!.class is (o.a = 1) if $!.class?

(Backwards compatibility might be a reasonable good reason, but I can't imagine it being useful.)

[shyouhei (Shyouhei Urabe)]

Martin Dürst wrote:

@matz (Yukihiro Matsumoto): Is there a good reason that o.a = 1 rescue $!.class is interpreted as o.a = (1 rescue $1.class), while it looks to me as if o.a = 1 if $!.class is (o.a = 1) if $!.class?

(Backwards compatibility might be a reasonable good reason, but I can't imagine it being useful.)

I guess issue #12402 is the reason behind this behaviour.

[shyouhei (Shyouhei Urabe)]

• Related to Bug #12402: Inline rescue behavior inconsistent for method calls with arguments and assignment added

[jeremyevans0 (Jeremy Evans)]

• Status changed from Open to Closed