Bug #20189: `rb_str_resize` does not clear coderange when expanding

• Status: Open
• Assignee: -
• Target version: -
• ruby -v: ruby 3.4.0dev (2024-01-09T07:07:19Z master db476cc71c) [x86_64-linux]
• Backport: 3.0: DONTNEED, 3.1: DONTNEED, 3.2: REQUIRED, 3.3: REQUIRED

[tompng (tomoya ishida)]

Expanding string in some encoding (utf16 utf32) can change coderange to either valid or broken, but rb_str_resize does not clear coderange.

This will cause a bug in c-extension libraries that use rb_str_resize.

# Example for stringio
s = StringIO.new("\0".encode('UTF-16LE'))

s.truncate(1); s.truncate(2); s.string.valid_encoding?
#=> true
s.truncate(1); s.string.valid_encoding?; s.truncate(2); s.string.valid_encoding?
#=> false (expect to be true)

[nobu (Nobuyoshi Nakada)]

Does this happen only with wide-char encoding?

[tompng (tomoya ishida)]

I think so. sjis char does not end with null bytes, other encoding seems same too.

Encoding.list.select {|e|
  256.times.any? do |first_byte|
    a = first_byte.chr
    b = a + "\0";
    # only one of \x??\x00 and \x?? is valid
    a.force_encoding(e).valid_encoding? != b.force_encoding(e).valid_encoding?
  end
}
# => [#<Encoding:UTF-16BE>, #<Encoding:UTF-16LE>]

It looks like there is no string like ("表"(sjis)=="\x95\x5c") that satisfies "\x??\x00" is valid and "\x??" is not.

I opened a pull request https://github.com/ruby/ruby/pull/9552

[nobu (Nobuyoshi Nakada)]

Seems b0b9f7201acab05c2a3ad92c3043a1f01df3e17f.

[nobu (Nobuyoshi Nakada)]

• Backport changed from 3.0: UNKNOWN, 3.1: UNKNOWN, 3.2: UNKNOWN, 3.3: UNKNOWN to 3.0: DONTNEED, 3.1: DONTNEED, 3.2: REQUIRED, 3.3: REQUIRED

[byroot (Jean Boussier)]

Expanding string in some encoding (utf16 utf32) can change coderange to either valid or broken,

I must admit I'm not very familiar with wide char encodings, but this surprises me a bit. Ruby strings should always have their terminator, so I don't see how expanding a string would change their interpretation.

[Eregon (Benoit Daloze)]

byroot (Jean Boussier) wrote in #note-5:

I must admit I'm not very familiar with wide char encodings, but this surprises me a bit. Ruby strings should always have their terminator, so I don't see how expanding a string would change their interpretation.

It's because in UTF-16 if the number of bytes is not a multiple of 2 then it's CR_BROKEN. Same for UTF-32 if not a multiple of 4.
And since rb_str_resize() changes the String#bytesize then that condition can change:

irb(main):002:0> "a".force_encoding(Encoding::UTF_16LE).valid_encoding?
=> false
irb(main):003:0> "a\x00".force_encoding(Encoding::UTF_16LE).valid_encoding?
=> true