Issue #16759 has been updated by Eregon (Benoit Daloze).

Assignee set to nobu (Nobuyoshi Nakada)

Could someone take a look at this place?
It seems a pretty serious bug to segfault/abort when using `Binding#local_variable_set`.
cc @nobu @usa

----------------------------------------
Bug #16759: MinGW 2.5 - SEGV bug with Binding#local_variable_set
https://bugs.ruby-lang.org/issues/16759#change-85918

* Author: MSP-Greg (Greg L)
* Status: Open
* Priority: Normal
* Assignee: nobu (Nobuyoshi Nakada)
* ruby -v: Ruby MinGW 2.5
* Backport: 2.5: UNKNOWN, 2.6: UNKNOWN, 2.7: UNKNOWN
----------------------------------------
A failure with MinGW 2.5.8 appeared in ruby/spec CI from the below in core/binding/local_variable_set_spec.rb.  I believe the specs passed on 2.5.7:
```ruby
it "raises a NameError on global access" do
  bind = binding
  -> { bind.local_variable_set(:$0, "") }.should raise_error(NameError)
end

it "raises a NameError on special variable access" do
  bind = binding
  -> { bind.local_variable_set(:$~, "") }.should raise_error(NameError)
  -> { bind.local_variable_set(:$_, "") }.should raise_error(NameError)
end
```

Working locally, the below code failed on both 2.5.7 and 2.5.8, but passed on mswin head & MinGW 2.6 thru head:
```ruby
[:$0, :$~, :$_].each do |v|
  a = -> (bind) do
    bind.local_variable_set(v, '')
  end
  begin
    a[binding]
  rescue => e
    puts "#{v}  #{e.class}"
  end
```

I checked other uses, see https://github.com/MSP-Greg/ruby-testing/blob/master/binding-local-variable.rb.  I ran the file in Actions CI on both Ubuntu & Windows, see the 'binding issue' step on any job at:
https://github.com/MSP-Greg/ruby-testing/actions/runs/71232731



-- 
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>