Three new mathn rubyspec failures on new specs

Hiro Asari added a comment - 08/Oct/09 12:05 AM Note that these should be really 1.9-only. The failing tests look like this: ((-8.0) ** (1/3.0)).should be_close(Complex(1, 1.73205), TOLERANCE) With 1.8.7, we get: surfboard:~$ ruby -v -e 'puts (-8)**(1/3.0)' ruby 1.8.7 (2008-08-11 patchlevel 72) [universal-darwin10.0] -e:1: warning: (...) interpreted as grouped expression NaN while with svn turnk, we get: surfboard:~$ ruby19 -v -e 'p (-8)**(1/3.0)' ruby 1.9.2dev (2009-10-08 trunk 25216) [x86_64-darwin10.0.0] -e:1: warning: (...) interpreted as grouped expression (1.0000000000000002+1.7320508075688772i) The attached patch partially solves this issue. surfboard:jruby[git:personal]$ bin/jruby --1.9 -v -e 'p (-8.0)**(1/3.0)' jruby 1.5.0dev (ruby 1.9.2dev trunk 24787) (2009-10-07 72d49b9) (Java HotSpot(TM) 64-Bit Server VM 1.6.0_15) [x86_64-java] -e:1 warning: (...) interpreted as grouped expression (1.0+1.73205080756888i) surfboard:jruby[git:personal]$ bin/jruby --1.9 -v -e 'p (-8)**(1/3.0)' jruby 1.5.0dev (ruby 1.9.2dev trunk 24787) (2009-10-07 72d49b9) (Java HotSpot(TM) 64-Bit Server VM 1.6.0_15) [x86_64-java] -e:1 warning: (...) interpreted as grouped expression (1.0+1.73205080756888i) surfboard:jruby[git:personal]$ bin/jruby --1.9 -v -e 'p (-0x8000_0000_0000)**(1/3.0)' jruby 1.5.0dev (ruby 1.9.2dev trunk 24787) (2009-10-07 72d49b9) (Java HotSpot(TM) 64-Bit Server VM 1.6.0_15) [x86_64-java] -e:1 warning: (...) interpreted as grouped expression (26007.978835447+45047.1407451702i) surfboard:jruby[git:personal]$ bin/jruby --1.9 -v -e 'p (-0x8000_0000_0000_0000-47)**(1/3.0)' jruby 1.5.0dev (ruby 1.9.2dev trunk 24787) (2009-10-07 72d49b9) (Java HotSpot(TM) 64-Bit Server VM 1.6.0_15) [x86_64-java] -e:1 warning: (...) interpreted as grouped expression -e:1: NaN (FloatDomainError) This is a partial fix, because: The spec won't run in 1.9 mode, since in 1.9, mathn wants to require 'mathn/complex' and 'mathn/rational', which are not implemented in JRuby yet. (This probably should be another ticket altogether.) The spec probably fails. The last invocation to Bignum#** should return a value (as MRI does).

Hiro Asari added a comment - 30/Oct/09 10:11 PM Since JRUBY-4072 is resolved, here are the updated results regarding the specs mentioned in this ticket: surfboard:jruby[git:personal]$ bin/jruby spec/mspec/bin/mspec -T--1.9 spec/ruby/library/mathn/fixnum/exponent_spec.rb spec/ruby/library/mathn/float/exponent_spec.rb spec/ruby/library/mathn/bignum/exponent_spec.rb jruby 1.5.0.dev (ruby 1.9.2dev trunk 24787) (2009-10-30 28b8827) (Java HotSpot(TM) 64-Bit Server VM 1.6.0_15) [x86_64-java] ....E. 1) Bignum#** returns self raised to other (positive) power ERROR NoMethodError: undefined method `power!' for 9223372036854775855:Bignum /Users/asari/Development/src/jruby/lib/ruby/1.9/mathn.rb:50:in `**' /Users/asari/Development/src/jruby/spec/ruby/library/mathn/bignum/exponent_spec.rb:10 /Users/asari/Development/src/jruby/spec/ruby/library/mathn/bignum/exponent_spec.rb:4 /Users/asari/Development/src/jruby/spec/ruby/library/mathn/bignum/exponent_spec.rb:55:in `load' Finished in 0.261000 seconds 3 files, 6 examples, 10 expectations, 0 failures, 1 error Note the new NoMethodError, which appears to be a result of the broken 'mathn.rb'. (This is fixed in http://svn.ruby-lang.org/cgi-bin/viewvc.cgi/trunk/lib/mathn.rb?view=markup&pathrev=25067 ). If you update mathn.rb, all the specs will pass—with a caveat that the spec for Bignum is probably not quite correct. That's Rubyspec's problem.

Hiro Asari added a comment - 14/Nov/09 2:22 PM The original patch modified the wrong code branch in RubyBignum.java. This one should be fine now: surfboard:~$ ruby1.9 -v; ruby1.9 -e 'p (-0x8000_0000_0000_0000-47)**(1/3.0)' ruby 1.9.2dev (2009-11-14 trunk 25765) [x86_64-darwin10.0.0] (1048575.9999999994+1816186.9075973416i) surfboard:jruby[git:personal]$ jruby --1.9 -e 'p (-0x8000_0000_0000_0000-47)**(1/3.0)' (1048576.0+1816186.90759734i)

Hiro Asari added a comment - 18/Nov/09 12:29 AM The new NoMethodError mentioned above is a separate issue: JRUBY-4247 .

Vladimir Sizikov added a comment - 19/Nov/09 4:02 PM I mean, at the very least we need some more specs that actually fail with JRuby, to make sure that the patch is actually fixes them.

Hiro Asari added a comment - 19/Nov/09 5:51 PM Vladimir, I think you're right. These specs pass without any alteration.

Vladimir Sizikov added a comment - 20/Nov/09 2:21 AM So, we're going to close this bug and open another one, and the new specs?

Hiro Asari added a comment - 20/Nov/09 3:24 AM Yes, I think we can close this ticket.

Vladimir Sizikov added a comment - 20/Nov/09 2:44 PM OK, closing this one. But, Hiro, please do add new specs since it seems that there is something that JRuby does differently from MRI. And then your patch would probably be handy.