Invalid UTF-8 Strings when compiling with MRB_UTF8_STRING compute length incorrectly

[lopopolo]

When compiling mruby with UTF-8 Strings (e.g. setting CFLAGS="-DMRB_UTF8_STRING"), mruby incorrectly computes the length of strings with invalid UTF-8 byte sequences.

mruby

$ git rev-parse HEAD
69482dbc8e590ed66f0944e9b48c4f9c2f83c873
$ git show
commit 69482dbc8e590ed66f0944e9b48c4f9c2f83c873 (HEAD -> master, origin/master, origin/HEAD)
Merge: 6587269a f7ff4810
Author: Yukihiro "Matz" Matsumoto <matz@ruby.or.jp>
Date:   Fri Jan 8 23:10:49 2021 +0900

    Merge pull request #5265 from shuujii/reapply-116e128b-because-it-is-back-at-456878ba

    Reapply 116e128b because it is back at 456878ba

Reproduction steps

rake clean
CFLAGS="-DMRB_UTF8_STRING" rake

Executing in mirb:

$ ./bin/mirb
mirb - Embeddable Interactive Ruby Shell

> xs = [192, 128].pack("C*")
 => "��"
> xs.bytes
 => [192, 128]
> xs.length
 => 1

Reference MRI execution

$ irb
[2.6.6] > xs = [192, 128].pack("C*")
=> "\xC0\x80"
[2.6.6] > xs.bytes
=> [192, 128]
[2.6.6] > xs.length
=> 2

With forced UTF-8 encoding:

$ irb
[2.6.6] > xs = [192, 128].pack("C*")
=> "\xC0\x80"
[2.6.6] > xs = xs.force_encoding(Encoding::UTF_8)
=> "\xC0\x80"
[2.6.6] > xs.encoding
=> #<Encoding:UTF-8>
[2.6.6] > xs.length
=> 2

[matz]

The length of broken UTF-8 strings is undefined, I think. You may get whatever length.
CRuby does validate if the string is UTF-8 compliant. mruby just assumes the string is valid UTF-8.
I will investigate if we can validate without hindering the performance.

[matz]

For this particular case (the first byte is in 80..C1) is checked now.