JRuby

Kernel#rand implementation does not match mri and rbx

Details

  • Type: Bug Bug
  • Status: Resolved Resolved
  • Priority: Major Major
  • Resolution: Fixed
  • Affects Version/s: None
  • Fix Version/s: JRuby 1.7.0.RC1
  • Component/s: None
  • Labels:
    None
  • Number of attachments :
    0

Description

this breaks certain test frameworks that assume calling srand will result in a specific repeatable sequence of random numbers.

it looks like there is a compatible random implementation in org.jruby.util.Random, but it is not being used.

$ /usr/bin/ruby -ve 'srand(1); p rand'
ruby 1.8.7 (2009-06-12 patchlevel 174) [universal-darwin10.0]
0.417022004702574

$ ruby -ve 'srand(1); p rand'
ruby 1.9.2p180 (2011-02-18 revision 30909) [x86_64-darwin10.6.0]
0.417022004702574

$ ruby -ve 'srand(1); p rand' 
rubinius 1.2.4dev (1.8.7 d9624587 yyyy-mm-dd JI) [x86_64-apple-darwin10.6.0]
0.417022004702574

$ ruby -ve 'srand(1); p rand'
jruby 1.6.0 (ruby-1.8.7-p330) (2011-04-05 2701790) (Java HotSpot(TM) 64-Bit Server VM 1.6.0_24) [darwin-x86_64-java]
0.730878190703291

Activity

Ascending order - Click to sort in descending order
Hide
Permalink
Aman Gupta added a comment -

fyi, the test framework in question was machinist-1.0.6 https://github.com/notahat/machinist/blob/v1.0.6/lib/sham.rb#L64-77

this is kind of a ridiculous bug, so feel free to mark as WontFix.

Show
Aman Gupta added a comment - fyi, the test framework in question was machinist-1.0.6 https://github.com/notahat/machinist/blob/v1.0.6/lib/sham.rb#L64-77 this is kind of a ridiculous bug, so feel free to mark as WontFix.
Hide
Permalink
David Calavera added a comment -

There is a branch with a compatible version of Random here:

https://github.com/jruby/jruby/tree/MersenneTwister

but I think nobody merged it yet

Show
David Calavera added a comment - There is a branch with a compatible version of Random here: https://github.com/jruby/jruby/tree/MersenneTwister but I think nobody merged it yet
Hide
Permalink
David Calavera added a comment -

At first glance it seems the MersenneTwister branch matches the other implementations:

[22:12:11][jruby-1.6.0][~/dev/jruby (MersenneTwister)]
$ bin/jruby -ve 'srand(1); p rand'
jruby 1.6.0.RC1 (ruby 1.8.7 patchlevel 330) (2011-04-06 fd957a1) (Java HotSpot(TM) 64-Bit Server VM 1.6.0_24) [darwin-x86_64-java]
0.417022004702574
Show
David Calavera added a comment - At first glance it seems the MersenneTwister branch matches the other implementations: [22:12:11][jruby-1.6.0][~/dev/jruby (MersenneTwister)] $ bin/jruby -ve 'srand(1); p rand' jruby 1.6.0.RC1 (ruby 1.8.7 patchlevel 330) (2011-04-06 fd957a1) (Java HotSpot(TM) 64-Bit Server VM 1.6.0_24) [darwin-x86_64-java] 0.417022004702574
Hide
Permalink
Aman Gupta added a comment -

Yep, it appears some of the commits already exist on master but not all of them. +1 for merging that fork.

Show
Aman Gupta added a comment - Yep, it appears some of the commits already exist on master but not all of them. +1 for merging that fork.
Hide
Permalink
Hiroshi Nakamura added a comment -

Sorry for not working actively these days. Related issue is at: JRUBY-5335

We talked that we'll merge it to master after releasing 1.6.0 for 1.7. At present, we don't have jruby_1_6 branch so merging it to master means that we'll push it for 1.6.1.

IIRC, Tom said that we should wait to create jruby_1_6 branch for a while, for concentrating 1.6.1 development. Guys, do you think it's time to create the branch for 1.6 series?

Show
Hiroshi Nakamura added a comment - Sorry for not working actively these days. Related issue is at: JRUBY-5335 We talked that we'll merge it to master after releasing 1.6.0 for 1.7. At present, we don't have jruby_1_6 branch so merging it to master means that we'll push it for 1.6.1. IIRC, Tom said that we should wait to create jruby_1_6 branch for a while, for concentrating 1.6.1 development. Guys, do you think it's time to create the branch for 1.6 series?
Hide
Permalink
Thomas E Enebo added a comment -

I think this desire to keep development for 1.6.x is not well-enough received. Unless someone objects I will cut a 1.6 branch tomorrow when we put out 1.6.1. This means lots of cherry-picking for us though.

Show
Thomas E Enebo added a comment - I think this desire to keep development for 1.6.x is not well-enough received. Unless someone objects I will cut a 1.6 branch tomorrow when we put out 1.6.1. This means lots of cherry-picking for us though.
Hide
Permalink
Charles Oliver Nutter added a comment -

Mersenne twister will be in 1.7. Marking as such.

Show
Charles Oliver Nutter added a comment - Mersenne twister will be in 1.7. Marking as such.
Hide
Permalink
Hiro Asari added a comment -

This can be marked Resolved, right?

Show
Hiro Asari added a comment - This can be marked Resolved, right?
Hide
Permalink
Charles Oliver Nutter added a comment -

I forget when exactly we got the fix in, but it's in. Marking for RC1.

Show
Charles Oliver Nutter added a comment - I forget when exactly we got the fix in, but it's in. Marking for RC1.

People

  • Assignee:
    Charles Oliver Nutter
    Reporter:
    Aman Gupta
Vote (0)
Watch (2)

Dates

  • Created:
    Updated:
    Resolved: