Bug #17419: `binding.eval` backtrace differente

• Status: Closed
• Priority: Normal
• Assignee: -
• Target version: -
• ruby -v: ruby 3.0.0rc1 (2020-12-20 master 8680ae9cbd) [x86_64-linux]
• Backport: 2.5: UNKNOWN, 2.6: UNKNOWN, 2.7: UNKNOWN

[vo.x (Vit Ondruch)]

Can somebody elaborate why there is different output for Ruby 2.7 vs Ruby 3.0?

$ ruby -v
ruby 2.7.2p137 (2020-10-01 revision 5445e04352) [x86_64-linux]

$ ruby -e '                                                                                                                                         
begin
binding.eval(%|raise "oops"|)
rescue Exception => exc
puts exc.backtrace
end
'
-e:3:in `<main>'
-e:3:in `eval'
-e:3:in `<main>'

$ ruby -v
ruby 3.0.0rc1 (2020-12-20 master 8680ae9cbd) [x86_64-linux]

$ ruby -e '                                                                                                                                         
begin
binding.eval(%|raise "oops"|)
rescue Exception => exc
puts exc.backtrace
end
'
(eval):1:in `<main>'
-e:3:in `eval'
-e:3:in `<main>'

IOW the first line of backtrace is different. Is this expected? This is causing test failures:

https://github.com/rails/web-console/issues/301

[vo.x (Vit Ondruch)]

• Related to Bug #17160: Wrong exception backtrace added

[shevegen (Robert A. Heiler)]

I think this is an intended change; it was explained via a link
to a comment in another issue (I think by mame or someone else,
where matz commented on reverting the behaviour). If I remember
the reason, the changed backtrace confused many ruby users over
the years, and it seemed better to go back to the state prior
to the change, hopefully lessening the confusion in the
process. :D

(Not sure if NEWS mentions this, but if not then perhaps a link
should be there, or just a short explanation in NEWS itself.)

[jeremyevans0 (Jeremy Evans)]

• Status changed from Open to Closed

This change is expected, see #4352.

[mame (Yusuke Endoh)]

Replacing b.eval(src) with b.eval(src, *b.source_location) will solve the issue.

However, #4352 focused on only Kernel#eval. I was not aware about the behavior change of Binding#eval. I've just added an explanation to NEWS.md at 4a18cc7f604c259415d933acd9fd00234733cab6.