Bug #13282: opt_str_freeze does not always dedupe

• Status: Rejected
• Assignee: -
• Target version: -
• Backport: 2.2: WONTFIX, 2.3: WONTFIX, 2.4: WONTFIX

[normalperson (Eric Wong)]

Attached is the same test extracted from test/ruby/test_optimization.rb
but it fails in a standalone context. Not sure why...

It's not a new breakage in trunk, either.

Testing with some old installs I had around, 2.1.9p490 seems fine,
however even Ruby 2.2.1 seems broken

[normalperson (Eric Wong)]

[e-mail] wrote:

Testing with some old installs I had around, 2.1.9p490 seems fine,

Too Bad I'm Wrong.

It's seems to be only tickled by the gemification of test-unit
in 2.2. Using the test-unit gem with 2.1 still fails...

Now, I wonder what's causing this to interact badly with
test-unit...

[normalperson (Eric Wong)]

Eric Wong [e-mail] wrote:

Now, I wonder what's causing this to interact badly with
test-unit...

Same test with minitest (5.10.1) + ruby 2.5.0dev r57744 works fine:

require 'minitest/autorun'
class TestString < Minitest::Test
def test_string_freeze
assert_equal "foo".freeze.object_id, "foo".freeze.object_id
end
end

Strange...

[normalperson (Eric Wong)]

Eric Wong [e-mail] wrote:

Now, I wonder what's causing this to interact badly with
test-unit...

power_assert/enable_tracepoint_events - this disables all
sorts of MRI optimizations for testing.

For background, I was relying on opt_str_freeze to detect
the presence of the optimization so I could enable (or skip)
further tests for my own optimizations which were mainline
Ruby-specific.

[ko1 (Koichi Sasada)]

• Status changed from Open to Rejected

for this case, specialized_instruction compilation option disable this feature.

[naruse (Yui NARUSE)]

• Backport changed from 2.2: REQUIRED, 2.3: REQUIRED, 2.4: REQUIRED to 2.2: WONTFIX, 2.3: WONTFIX, 2.4: WONTFIX