Hash has bug to set encoding into key string wrongly when the key string is used once with different encoding #3405
Description
On JRuby 9.0.1.0, Hash sets wrong encoding only when the key string is used as Hash key with different encoding.
hash1 = {}
hash1['str'.force_encoding('ASCII-8BIT')] = 1
p hash1.keys.first.encoding # ASCII-8BIT
hash2 = {}
hash2['str'.force_encoding('UTF-8')] = 1
p hash2.keys.first.encoding # ASCII-8BIT !? (expected: UTF-8)Script to show situations to reproduce bugs:
# encoding: ascii-8bit
str = 'hello'
obj1 = {'hello' => 1}
obj2 = {str => 2}
p({str: str, enc: str.encoding, k1enc: obj1.keys.first.encoding, k2enc: obj2.keys.first.encoding})
str = 'hello'.force_encoding('UTF-8')
obj1 = {'hello'.force_encoding('UTF-8') => 1}
obj2 = {str => 2}
p({str: str, enc: str.encoding, k1enc: obj1.keys.first.encoding, k2enc: obj2.keys.first.encoding})
str = 'hello'.force_encoding('UTF-8')
obj1 = {}
obj1[str] = 1
obj2 = {}
obj2[str] = 2
p({str: str, enc: str.encoding, k1enc: obj1.keys.first.encoding, k2enc: obj2.keys.first.encoding})
str = 'world'.force_encoding('UTF-8')
obj1 = {'world'.force_encoding('UTF-8') => 1}
obj2 = {str => 2}
p({str: str, enc: str.encoding, k1enc: obj1.keys.first.encoding, k2enc: obj2.keys.first.encoding})
str = 'unused'
obj1 = {'unused'.force_encoding('UTF-8') => 1}
obj2 = {str.force_encoding('UTF-8') => 2}
p({str: str, enc: str.encoding, k1enc: obj1.keys.first.encoding, k2enc: obj2.keys.first.encoding})Expected result is:
- First line: all keys are encoded in ASCII-8BIT
- Rest: all keys are encoded in UTF-8
Actual result in JRuby 9.0.1.0 is:
- First line: all keys are encoded in ASCII-8BIT
- All
helloas hash keys are encoded in ASCII-8BIT