Bug #20185: String#ascii_only? buggy in ruby 3.3

• Status: Closed
• Assignee: -
• Target version: -
• ruby -v: 3.3.0
• Backport: 3.0: UNKNOWN, 3.1: UNKNOWN, 3.2: UNKNOWN, 3.3: UNKNOWN

[chucke (Tiago Cardoso)]

This was the smallest reduction of the bug I could come up with:

require "stringio"

puts StringIO::VERSION

def is_ascii(buffer)
  str = buffer.string
  puts "\"#{str}\" is ascii: #{str.ascii_only?}"
end

buffer = StringIO.new("".b)

buffer.write("a=b&c=d")
buffer.rewind
is_ascii(buffer)
buffer.write "богус"
is_ascii(buffer)

# in ruby 3.3
#=> 3.1.0
#=> "a=b&c=d" is ascii: true
#=> "богус" is ascii: true

# in ruby 3.2
#=> 3.0.4
#=> "a=b&c=d" is ascii: true
#=> "богус" is ascii: true

# in ruby 3.1
#=> 3.0.1
#=> "a=b&c=d" is ascii: true
#=> "богус" is ascii: false

I believe that only the 3.1 result is correct, as "богус" first character is not ascii.

[andrykonchin (Andrew Konchin)]

I cannot reproduce the issue with plain String (without StringIO) on Ruby 3.2, 3.1 and 3.0. ascii_only? reports false for "богус":

ruby -e 'p "богус".ascii_only?'
false

I believe in the examples involving StringIO the observed behaviour is caused by preserving StringIO#string's encoding. StringIO instance is initialised with a String literal in binary encoding. And any modification like writing doesn't change encoding even when a UTF-8 String is written:

io = StringIO.new "".b
io.string.encoding # => #<Encoding:ASCII-8BIT>

io.write "汉"
io.string.encoding # => #<Encoding:ASCII-8BIT>

In case of the "богус" String literal there are bytes greater than 127 so they are treated as non-ASCII:

io = StringIO.new "".b
io.write "богус"
io.string.bytes # => [208, 177, 208, 190, 208, 179, 209, 131, 209, 129]

[nobu (Nobuyoshi Nakada)]

• Status changed from Open to Closed

[chucke (Tiago Cardoso)]

nobu, can I ask why was the ticket closed? Even considering the comment from andrykonchin, he clearly points oot at the end that there are bytes greater than 128 in the string (therefore .ascii_only? should be false).

[jeremyevans0 (Jeremy Evans)]

chucke (Tiago Cardoso) wrote in #note-3:

nobu, can I ask why was the ticket closed? Even considering the comment from andrykonchin, he clearly points oot at the end that there are bytes greater than 128 in the string (therefore .ascii_only? should be false).

This was fixed by 6283ae8d369bd2f8a022bb69bc5b742c58529dec

[chucke (Tiago Cardoso)]

Apologies everyone, got temporary redmine visual impairment. Thank you.

[Eregon (Benoit Daloze)]

Indeed on Redmine I see no link to the commit in https://bugs.ruby-lang.org/issues/20185?tab=history#note-2, it seems like a bug.

[hsbt (Hiroshi SHIBATA)]

No, Fix https://bugs.ruby-lang.org/issues/20185of commit message is not correct format for redmine autolink.