Using a string that has been mutated seems to result in a variable containing an unexpected value #8861
Description
Environment Information
- jruby 10.0.0.1 (3.4.2) 2025-05-07 79cf1e4 Java HotSpot(TM) 64-Bit Server VM 24.0.1+9-30 on 24.0.1+9-30 +indy +jit [x86_64-linux]
- This problem ONLY occurs if I run: JRUBY_OPTS="--debug" rspec
- Linux6.12.27-amd64
#1SMP PREEMPT_DYNAMIC Debian 6.12.27-1 (2025-05-06) x86_64 GNU/Linux
Expected Behavior
This one is hard to explain, but I'll try!
I have a piece of code that looks like this:
module_symbol = Util.classify(type_symbol).to_symand if I use JRUBY_OPTS="--debug" it winds up with a surprising value.
If I binding.pry at the right spot, the problem can be demonstrated like this:
[1] pry(Foobara::BuiltinTypes)> type_symbol
=> :attributes
[2] pry(Foobara::BuiltinTypes)> Util.classify(type_symbol)
/home/miles/.rbenv/versions/jruby-10.0.0.1/lib/ruby/gems/shared/gems/foobara-util-0.0.12/lib/foobara/util/string.rb:22: warning: literal string will be frozen in the future, the string was created here: /home/miles/.rbenv/versions/jruby-10.0.0.1/lib/ruby/gems/shared/gems/foobara-util-0.0.12/lib/foobara/util/string.rb:16
=> "stringintegerstringno_cast_needed_if_is_a_big_decimalintegerstringintegerstringno_cast_needed_if_is_a_stringnumericsymboldowncasematchesmax_lengthnumericsymbolno_cast_needed_if_is_a_datestringhashstringhashno_cast_needed_if_is_a_timeseconds_since_epochstringhashdateseconds_since_epochstringhashdateno_cast_needed_if_is_a_false_class_or_true_classnumericstring_or_symbolnumericstring_or_symboldowncasemust_have_an_at_signfirst_part_has_bad_characterssecond_part_has_bad_charactersfirst_part_cannot_start_with_or_end_with_a_dot_or_have_two_dots_in_a_rowsecond_part_cannot_be_emptydomain_cannot_start_with_or_end_with_a_hyphencannot_exceed64_characterscannot_have_multiple_at_signsfirst_part_cannot_be_emptydowncasemust_have_an_at_signfirst_part_has_bad_characterssecond_part_has_bad_charactersfirst_part_cannot_start_with_or_end_with_a_dot_or_have_two_dots_in_a_rowsecond_part_cannot_be_emptydomain_cannot_start_with_or_end_with_a_hyphencannot_exceed64_characterscannot_have_multiple_at_signsfirst_part_cannot_be_emptyno_cast_needed_if_is_a_arrayarrayablesizeelement_type_declarationarrayableelement_type_declarationsno_cast_needed_if_is_a_hasharraykey_type_declarationvalue_type_declarationarrayAttributesAttributes"
So type_symbol is :attributes and so the expected output of this utility function is "Attributes" but instead it looks like some unexpected value based on some object somewhere else in memory.
Note that warning about frozen strings. That's a clue of how to work around this problem.
Here's the .classify method, which works fine without JRUBY_OPTS="--debug":
def classify(string)
camelize(string, true)
end
def camelize(string, upcase_first = false)
return nil if string.nil?
if string.is_a?(::Symbol)
string = string.to_s
end
retval = ""
string.each_char do |char|
if ["_", "-"].include?(char)
upcase_first = true
elsif upcase_first
retval << char.upcase
upcase_first = false
else
retval << char.downcase
end
end
retval
end
Notice that line retval = "". If I change this to: retval = +"" then the problem goes away.
If really needing to reproduce this, it's in the jruby branch of https://github.com/foobara/foobara It's kind of a lot of steps but here's what I would try to do to reproduce it from scratch on a different machine:
git clone git@github.com:foobara/foobara
cd foobara
git checkout jruby
bundle
JRUBY_OPTS="--debug" rspec