This is a reaction to a [bug] we hit in web-console. The cause of it was
a `Kernel` extension called `#console` that was public and was fighting
over Railties with console block to be run on `rails console`. We solved
it by making the method private. We did that through `module_function`
so `::Kernel.console` can be invoked even in `BasicObject`.
I'm proposing to make most of the core Active Support `Kernel`
extensions `module_function` as well. Those are currently public and we
are polluting every `Object` public interface with them.
```ruby
>> Object.new.respond_to? :silence_warnings
=> true
>> Object.new.respond_to? :with_warnings
=> true
>> Object.new.respond_to? :enable_warnings
=> true
>> Object.new.respond_to? :suppress
=> true
``
Some extensions like `Kernel#class_eval` should be public, but most of
them don't really need to be.
[bug]: https://github.com/rails/web-console/issues/184
If you set the DATABASE_URL environment variable to `mydatabase` by accident, you end up getting a series of errors that are hard to trace. For example:
```
warning: already initialized constant ActiveRecord::Base::OrmAdapter
```
Turns out the cascade of errors is due to the error raised by `.tr` being called on `nil`.
This commit makes sure that `scheme` is set before calling `.tr` on it. My previous iteration used `@uri.scheme.try(:tr, '-', '_')` but using the `&&` logical operator is a fair bit faster: http://stackoverflow.com/questions/26655032/try-vs-performance
With this change, the error message becomes much more understandable:
```
FATAL: database "mydatabase" does not exist (ActiveRecord::NoDatabaseError)
```
Because of the expanding whitelist for test filters, this test ended up
running the tests on lines 4 and 9 in the post test even though the path
wasn't right.
Happened incidentally because the same line numbers were used in both
account and post test.
Add the .rb line so the file is required correctly and the filters are
applied.
`derive_regexp` was written with the assumption that we were run from a
blank slate — that if the filter didn't match we might as well return it
because it was nil.
This isn't the case because minitest calls `run` on every runnable. Which
is any subclass of Minitest::Runnable, such as ActiveSupport::TestCase,
ActionDispatch::IntegrationTest as well as any inheriting from those.
Thus after the first `run` we'd have put in a composite filter in
`options[:filter]` making the next `run` create a linked list when it
failed to match the regexp and put the composite filter as the head.
Every runnable would accumulate more and more of the same filters,
which effectively acted like an expanding whitelist and we ran tests
from other runnables.
Clog the accumulation by returning nil if there's no filter to derive
a regexp from.
Note: we pass a seed in the tests because Minitest shuffles the runnables
to ensure the whitelist is expanded enough that the failure is triggered.
Prathamesh Sonpatki