Multibyte characters preceding offense breaks position and formatting on cached results

[ghost]

When there are multibyte UTF-8 characters that precede an offense, then the position of offense and subsequent formatting and underlining is incorrect when displaying cached results.

Expected behavior

Cached results should match first-time results.

Actual behavior

Cached results do not match first-time results.

Steps to reproduce the problem

Run rubocop twice on a file containing this UTF-8 text:

puts "føø"
useless = 'bar'

Note that foo is spelled using the Unicode character U+00F8 'Latin Small Letter o With Stroke', which in UTF-8 requires two octets. The first time rubocop runs, it will correctly underline 'useless' and indicate a position of 2:1. On the second time, when it is using cached results, it underlines the end of the first line and indicates a position of 1:12.

RuboCop version

$ rubocop -V
0.39.0 (using Parser 2.3.0.7, running on ruby 2.3.0 x86_64-darwin15)

[owst]

This appears to have been fixed - running the HEAD version as of today (624705384ce42e529c479c10f03d0e99a5ae39f8) does not reproduce the issue. I think it was quite possibly fixed by #3000.

[owst]

@bbatsov I think this can be closed