Fix bug in RangeHelp where range_by_whole_lines goes past end of buffer

[EiNSTeiN-]

This fixes a bug where RangeHelp#range_by_whole_lines may return a range that goes past the end of the buffer when include_final_newline argument is true. This is because of the line end_offset += 1 if include_final_newline.

I added 2 tests that fail without this fix, the original bug I encountered was from BracesAroundHashParameters when the hash being removed occurs at the very end of the ruby code.

[EiNSTeiN-]

To add some more context, without this fix, an exception is raised when BracesAroundHashParameters attempts to auto-correct a problematic file.

     IndexError:
       The range file.rb:3:1 is outside the bounds of the source
     # /Users/einstein/.gem/ruby/2.3.3/gems/parser-2.5.1.0/lib/parser/source/tree_rewriter.rb:276:in `check_range_validity'
     # /Users/einstein/.gem/ruby/2.3.3/gems/parser-2.5.1.0/lib/parser/source/tree_rewriter.rb:268:in `combine'
     # /Users/einstein/.gem/ruby/2.3.3/gems/parser-2.5.1.0/lib/parser/source/tree_rewriter.rb:129:in `replace'
     # /Users/einstein/.gem/ruby/2.3.3/gems/parser-2.5.1.0/lib/parser/source/tree_rewriter.rb:153:in `remove'
     # /Users/einstein/src/github.com/bbatsov/rubocop/lib/rubocop/cop/corrector.rb:75:in `remove'
     # /Users/einstein/src/github.com/bbatsov/rubocop/lib/rubocop/cop/style/braces_around_hash_parameters.rb:146:in `remove_braces_with_range'
     # /Users/einstein/src/github.com/bbatsov/rubocop/lib/rubocop/cop/style/braces_around_hash_parameters.rb:133:in `remove_braces_with_whitespace'
     # /Users/einstein/src/github.com/bbatsov/rubocop/lib/rubocop/cop/style/braces_around_hash_parameters.rb:66:in `block in autocorrect'
     # /Users/einstein/src/github.com/bbatsov/rubocop/lib/rubocop/cop/corrector.rb:61:in `block (2 levels) in rewrite'
     # /Users/einstein/.gem/ruby/2.3.3/gems/parser-2.5.1.0/lib/parser/source/tree_rewriter.rb:220:in `transaction'
     # /Users/einstein/src/github.com/bbatsov/rubocop/lib/rubocop/cop/corrector.rb:60:in `block in rewrite'
     # /Users/einstein/src/github.com/bbatsov/rubocop/lib/rubocop/cop/corrector.rb:58:in `each'
     # /Users/einstein/src/github.com/bbatsov/rubocop/lib/rubocop/cop/corrector.rb:58:in `rewrite'
     # /Users/einstein/src/github.com/bbatsov/rubocop/lib/rubocop/cop/team.rb:126:in `autocorrect_all_cops'
     # /Users/einstein/src/github.com/bbatsov/rubocop/lib/rubocop/cop/team.rb:70:in `autocorrect'
     # /Users/einstein/src/github.com/bbatsov/rubocop/lib/rubocop/cop/team.rb:98:in `block in offenses'
     # /Users/einstein/src/github.com/bbatsov/rubocop/lib/rubocop/cop/team.rb:115:in `investigate'
     # /Users/einstein/src/github.com/bbatsov/rubocop/lib/rubocop/cop/team.rb:94:in `offenses'
     # /Users/einstein/src/github.com/bbatsov/rubocop/lib/rubocop/cop/team.rb:44:in `inspect_file'

[rrosenblum]

I assume that this wasn't working before the change. It seems like this is what should be in the CHANGELOG. We normally only add CHANGELOG references for changes that could impact users. The average user doesn't know what RangeHelp or range_by_whole_lines is.

[EiNSTeiN-]

Sounds good to me, do you have a preferred wording for the changelog?

[EiNSTeiN-]

I went with

Fix `Style/BracesAroundHashParameters` auto-correct sometimes going past the end of the file.

Does that work better?

[rrosenblum]

That sounds fine to me. Thanks for adding some context. Do you know what in particular about this line of code caused it go past the end of the file? Maybe something about the spacing or new lines?

[EiNSTeiN-]

the example broken source is on line 267

src = "render 'foo', {\n  foo: bar\n}"

the problem is that this cop wants to remove the braces around hash argument, and to do that it will try to select the entire line where } is located including a possible trailing newline, to do so range_by_whole_lines will add 1 character to the end of the line here:
https://github.com/bbatsov/rubocop/blob/a66faa88ce004f8717aff136230d196e8be3ee2c/lib/rubocop/cop/mixin/range_help.rb#L68
and this makes the range go past the end of the buffer because } in this example is the last character in the buffer.

[rrosenblum]

It sounds like the issue comes from a closing curly bracket, } living on a separate line than the contents of the hash. If that is the case, it would be useful to add to the CHANGELOG.

[EiNSTeiN-]

Added some clarification to changelog 👍

[bbatsov]

Thanks!