Issue #15926 has been updated by jeremyevans0 (Jeremy Evans).

File str-uminus-freeze-fix.patch added

I agree that this is a bug.  I'm not sure if `rb_fstring`'s behavior should be changed (hopefully a more knowledgeable committer can weigh in), so a conservative approach to fix this is to dup the receiver before passing to `rb_fstring`, if the receiver's class is not `String` and the object is not already frozen.  The attached patch does this.  

----------------------------------------
Bug #15926: Edge case issue with String#uminus
https://bugs.ruby-lang.org/issues/15926#change-78688

* Author: luke-gru (Luke Gruber)
* Status: Open
* Priority: Normal
* Assignee: 
* Target version: 
* ruby -v: 
* Backport: 2.4: UNKNOWN, 2.5: UNKNOWN, 2.6: UNKNOWN
----------------------------------------
I was working on issue related to code in `rb_fstring`(https://github.com/ruby/ruby/pull/2233) and saw some weird behavior in the function,
freezing the given string if it's not a "bare" string and it's small enough to be embedded.

The issue comes up in the following edge case:

class MyString < String
end
non_frozen = MyString.new("nonfrozen")
frozen = -non_frozen # deduplicates, but shouldn't freeze receiver
non_frozen << " added" # raises FrozenError

I'm not sure what the correct behavior should be with a subclass and String#uminus. Should it return
a frozen regular String or a frozen copy of the given class's string?

Not a practical concern (not often come upon I'm sure), but I think a valid one.

---Files--------------------------------
str-uminus-freeze-fix.patch (1.38 KB)


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

Unsubscribe: <mailto:ruby-core-request / ruby-lang.org?subject=unsubscribe>
<http://lists.ruby-lang.org/cgi-bin/mailman/options/ruby-core>