Ensure let block is not passed to super call.
I think this should fix it.
- Bundler 1.3 needs rbconfig to be required to run `bin/rspec` after installing with `--standalone --binstubs`. bundler/bundler#2382 - The Travis VMs no longer have 32-bit JVMs available. Aruba 0.4.11 forced the -d32 flag. 0.5 just documents it and makes a file available to require that will set it. We've updated Aruba here so that the jruby ENV vars aren't set automatically; we've kept the flag to disable JIT but removed -d32. - Aruba now has a "the file xyz shoudl contain" step.
Fix README example spec to display actual output
4.6.2 will no longer install on travis because it requires rubygems ~> 1.8, and travis now has rubygems 2.0 installed.
Fix #820. This is directly related to bug http://bugs.ruby-lang.org/issues/8059. It seems to affect Ruby 1.9.3 but not Ruby 2.0.0. As per the notes, it is a heisenbug, so simply looking at it causes the bug to disappear. Force the call to super to make sure the let block is not passed in.
Revise before(:all)/let behavior based on feedback from @dchelimsky.
- This is closer to the behavior we've had in 2.12 and earlier. `let` declarations that are accessed in `before(:all)` will preserve their memoized value for all examples in the group. - On 2.12 and earlier, this would cause additional `let` declarations that were not accessed in `before(:all)` but were accessed in individual examples to "leak", because the same `__@memoized` ivar got shared with each example in the group, and would be mutated each time an additional `let` declaration was accessed. This commit changes that behavior. The fact that one `let` declaration is accessed in `before(:all)` does not affect the lifecycle of other `let` declarations in any way.
Include caller in warning msg for `let` usage in `before(:all)`.
Clear let/subject memoization when referenced in before(:all).
...since `mock` and `stub` have been deprecated in rspec-mocks.
Candidate fix for #800
Require set where it's needed. Fixes #808
Uses secure RubyGems URL
- Ensure the NamedSubject override module gets inserted in the ancestor chain at the right place (this wasn't guaranteed before). - Raise an error when `super` is used in a named subject. I don't think we should support it.
Fixes #800. Note that the complexity of supporting `super` in a named subject makes me think that we should disable `super` in named subjects altogether instead.
- The options state really belongs with the hook when the instance is created. Better to pass it in the initializer than to have a separate `with` method for setting it. - No need to wrap/unwrap the block with a proc an extra time; we can just pass the proc directly into the initializer.