Bug #18624: `const_source_location` returns [false, 0] when autoload is defined for the constant

• Status: Closed
• Assignee: -
• Target version: -
• ruby -v: ruby 3.2.0dev (2022-03-11T08:38:13Z master 2e4516be26) [x86_64-linux]
• Backport: 2.6: UNKNOWN, 2.7: UNKNOWN, 3.0: UNKNOWN, 3.1: UNKNOWN

[pocke (Masataka Kuwabara)]

const_source_location returns [false, 0] unexpectedly with the following code.

# test.rb
path = File.join(__dir__, 'test2')
Object.autoload 'Test2', path
require path

p Object.const_source_location 'Test2'

# test2.rb
class Test2
end

$ ruby -v test.rb
ruby 3.2.0dev (2022-03-11T08:38:13Z master 2e4516be26) [x86_64-linux]
[false, 0]

I expect it returns ['/path/to/test2.rb', 2], but it doesn't.
Ruby 2.7 ~ 3.1 have the same behavior.

I actually encountered this problem in a Rails app with Zeitwerk. The following codes are a reproducing example with Zeitwerk.
loader.setup calls autoload, so it causes the same problem.

# test.rb
require "zeitwerk"
loader = Zeitwerk::Loader.for_gem
loader.setup

require File.join(__dir__, 'test2')

p Zeitwerk::VERSION
p Object.const_source_location 'Test2'

# test2.rb
class Test2
end

[pocke (Masataka Kuwabara)]

• Description updated (diff)

[jeremyevans0 (Jeremy Evans)]

I've submitted a pull request to fix this: https://github.com/ruby/ruby/pull/5646

[ufuk (Ufuk Kayserilioglu)]

This is a duplicate of Issue #17354 and @jeremyevans0 (Jeremy Evans), in that issue, you had argued that just because the autoload location was known didn't mean that the constant would be guaranteed to be defined by that file.

I believe this was also discussed in the Dev Meeting at the time and a consensus wasn't reached: https://bugs.ruby-lang.org/issues/17354#note-12

I suggest we close this issue and continue the conversation on #17354

[jeremyevans0 (Jeremy Evans)]

ufuk (Ufuk Kayserilioglu) wrote in #note-3:

This is a duplicate of Issue #17354 and @jeremyevans0 (Jeremy Evans), in that issue, you had argued that just because the autoload location was known didn't mean that the constant would be guaranteed to be defined by that file.

I believe this was also discussed in the Dev Meeting at the time and a consensus wasn't reached: https://bugs.ruby-lang.org/issues/17354#note-12

I suggest we close this issue and continue the conversation on #17354

Sorry, but you are incorrect. #17354 is regarding behavior of const_source_location before the autoloaded file is required. This issue is regarding behavior of const_source_location after the autoloaded file is required. They are separate issues, not duplicates, and should be treated as such. This issue is definitely a bug, because the related constant has already been defined and the source of the constant is now known.

[ufuk (Ufuk Kayserilioglu)]

jeremyevans0 (Jeremy Evans) wrote in #note-4:

Sorry, but you are incorrect. #17354 is regarding behavior of const_source_location before the autoloaded file is required. This issue is regarding behavior of const_source_location after the autoloaded file is required. They are separate issues, not duplicates, and should be treated as such. This issue is definitely a bug, because the related constant has already been defined and the source of the constant is now known.

Indeed, thanks for the correction. I had missed the require path line in the original reproduction example. Apologies for the noise and thanks again for the clarification.

• Status changed from Open to Closed

Applied in changeset git|c85d1cda86d75ee2c3f7b42f22c543409cb5a186.

---

Fix Module#const_source_location for autoload constants with direct requires

If an autoload exists for a constant, but the path for the autoload
was required, const_source_location would return [false, 0] instead
of the actual file and line. This fixes it by setting the appropriate
file and line in rb_const_set, and saving the file and line in
const_tbl_update before they get reset by current_autoload_data.

Fixes [Bug #18624]