[ruby-core:122817] [Ruby Bug#21517] Behaviour of StringIO#readpartial with multibyte differs from IO#readpartial

18 Jul 2025

      Issue #21517 has been reported by herwin (Herwin W).

----------------------------------------
Bug #21517: Behaviour of StringIO#readpartial with multibyte differs from IO#readpartial
https://bugs.ruby-lang.org/issues/21517

* Author: herwin (Herwin W)
* Status: Open
* ruby -v: ruby 3.4.5 (2025-07-16 revision 20cda200d3) +PRISM [x86_64-linux] (+ others)
* Backport: 3.2: UNKNOWN, 3.3: UNKNOWN, 3.4: UNKNOWN
----------------------------------------
The specs for `IO#readpartial` have the following test (https://github.com/ruby/spec/blob/ed254bae321221f5cd24280beb37e5bddaff6bb6/c...):
```ruby
before :each do
  @rd, @wr = IO.pipe
  @rd.binmode
  @wr.binmode
end

it "reads after ungetc with multibyte characters in the buffer" do
  @wr.write("∂φ/∂x = gaîté")
  c = @rd.getc
  @rd.ungetc(c)
  @rd.readpartial(3).should == "\xE2\x88\x82"
  @rd.readpartial(3).should == "\xCF\x86/"
end
```

This behaviour is different in `StringIO#readpartial`. If I add this as a spec:
```ruby
it "reads after ungetc with multibyte characters in the buffer" do
  @string = StringIO.new(+"∂φ/∂x = gaîté")
  c = @string.getc
  @string.ungetc(c)
  @string.readpartial(3).should == "\xE2\x88\x82"
  @string.readpartial(3).should == "\xCF\x86/"
end
```
it fails with every supported Ruby version, and even with 3.5-dev of today, with the message `Expected "\xE2\x88\x82" == "∂"`

The `getc` + `ungetc` can be removed, but this does not change the behaviour.

-- 
https://bugs.ruby-lang.org/