Browse files

Fix the build

  • Loading branch information...
1 parent b7b37f7 commit a7e1ca21c9de24743f3c6f840b96711c0baa0890 @sferik sferik committed Jul 27, 2012
Showing with 1 addition and 13 deletions.
  1. +0 −4 .travis.yml
  2. +1 −9
4 .travis.yml
@@ -3,10 +3,6 @@ matrix:
- rvm: ruby-head
- - rbx-18mode
- - rbx-19mode
- - jruby-18mode
- - jruby-19mode
- 1.8.7
- 1.9.2
- 1.9.3
@@ -98,19 +98,11 @@ end
## Supported Ruby Versions
This library aims to support and is [tested against][travis] the following Ruby
* Ruby 1.8.7
* Ruby 1.9.2
* Ruby 1.9.3
-* [JRuby][]
-* [Rubinius][]
-If something doesn't work on one of these interpreters, it should be considered
-a bug.
This library may inadvertently work (or seem to work) on other Ruby
implementations, however support will only be provided for the versions listed

5 comments on commit a7e1ca2

octokit member

Fix the build

For various definitions of "fix". Why not support JRuby/Rubinius?

octokit member

If you look at the last failed build, the failures have absolutely nothing to do with our code. Our specs aren't even being run. The builds were crapping out before they got to that step.

As I see it, Ruby is defined by MRI. Yes, there's RubySpec, but that was reverse-engineered from MRI, including specifying long-standing MRI bugs.

If our specs pass on MRI but fail on JRuby, that's a bug in JRuby, not a bug in our code, so it shouldn't affect our overall build status. I suppose I could have moved JRuby and Rubinius to the allow_failures array so they don't affect our overall build status, but IMHO, the purpose of CI is not for us to discover bugs in JRuby and Rubinius.

If you asked me the same question 12 months ago, I probably would have given you a different answer. Since then, I've grown tired of investigating failures that I did not cause and am not responsible for fixing.

octokit member

The failures have been intermittent and random, too. I had chalked it up to Travis instability. @sferik, are you seeing the same on Twitter?

octokit member

Travis is partially to blame but the fact they keep breaking leads be to think the core problem is instability in JRuby and Rubinius. They simply shouldn't be that hard to get running.

I've kept these implementations enabled for the twitter gem for now, but I'm inclined to rip them out if I start seeing repeated failures.

octokit member

If our specs pass on MRI but fail on JRuby, that's a bug in JRuby, not a bug in our code

Fair enough; your commit message just threw me ;)

Please sign in to comment.