Move remaining core specs to use fewer shared examples

[Earlopain]

It's quite chonky but mostly more of the same for #1364. This converts the entirety of core specs.

Before:
3805 files, 35698 examples, 273269 expectations, 0 failure, 0 errors, 0 tagged

After:
3805 files, 34586 examples, 256944 expectations, 0 failure, 0 errors, 0 tagged

1112 fewer specs.

[Copilot]

Copilot wasn't able to review this pull request because it exceeds the maximum number of files (300). Try reducing the number of changed files and requesting a review from Copilot again.

[Earlopain]

Let me split this up in two PRs then

[Copilot]

Pull request overview

Copilot reviewed 150 out of 150 changed files in this pull request and generated 4 comments.

[Copilot]

Pull request overview

Copilot reviewed 150 out of 150 changed files in this pull request and generated 4 comments.

[Copilot]

Pull request overview

Copilot reviewed 150 out of 150 changed files in this pull request and generated 1 comment.

Comments suppressed due to low confidence (5)

core/io/tty_spec.rb:1

• This spec can be flaky across CI/container environments because /dev/tty may be missing or inaccessible (e.g., Errno::ENOENT, Errno::EACCES, Errno::ENODEV), not just Errno::ENXIO. Consider broadening the rescued errors (or guarding by checking existence/permissions) so the spec reliably skips when TTY access is not available.
  core/file/to_path_spec.rb:1
• This platform_is/guard block defines before/after hooks but contains no examples that use @dir. Either add the missing TMPFILE-related examples within this scope or remove the unused hooks to avoid dead setup code.
  core/float/fdiv_spec.rb:1
• The second example’s description duplicates the first one verbatim, which makes failures harder to interpret. Rename the second description to reflect the distinct intent (e.g., division by a very large Integer).
  core/kernel/dup_spec.rb:1
• The spec that verified #dup calls #initialize_copy was removed, which drops coverage for an important and observable part of Ruby’s object-copying semantics. Consider re-adding an explicit example asserting initialize_copy is invoked (or adding equivalent coverage elsewhere in the same spec suite) so regressions in dup hooks are caught.
  core/kernel/clone_spec.rb:1
• Similarly to dup_spec, removing the initialize_copy expectation for #clone reduces coverage for clone’s copy hook behavior. Re-introduce an example that asserts initialize_copy is called during clone to keep coverage of this user-visible contract.

[eregon]

Great stuff and some of the deduplication notably for Marshal will be very nice when improving those specs.

[eregon]

For Float we have inspect is alias of to_s, I think that's good because to_s is more frequent than inspect, could you follow the pattern used for Float for other classes then?

[Copilot]

Pull request overview

Copilot reviewed 151 out of 151 changed files in this pull request and generated no new comments.

Comments suppressed due to low confidence (5)

core/method/to_s_spec.rb:1

• Method#to_s is being tested using the :inspect method symbol. This makes the spec exercise the wrong API and will mask regressions in Method#to_s. Update the shared example invocations to use :to_s here.
  core/fiber/scheduler_spec.rb:1
• This example has no assertions and will always pass, effectively removing coverage for Fiber.scheduler while still reporting success. Either remove this spec file entirely (if intentionally redundant) or mark it as skipped (so it doesn't appear as a passing test), or add a minimal assertion that Fiber.scheduler returns the currently configured scheduler.
  core/kernel/dup_spec.rb:1
• The spec that verified dup calls #initialize_copy was removed, and there is no replacement in this file. Since initialize_copy is part of the observable contract of dup, this looks like a coverage regression; consider restoring the removed example or adding an equivalent assertion elsewhere.
  core/kernel/clone_spec.rb:1
• Similar to dup_spec, the spec that verified clone calls #initialize_copy was removed without an apparent replacement. Consider restoring this coverage (or relocating it) to avoid losing behavioral guarantees around clone.
  core/kernel/to_enum_spec.rb:1
• The expectation here is hard to read because the RHS uses Array#each with a block (which returns the receiver array), while the LHS uses Enumerator#map (which allocates a new array). Consider rewriting the assertion to compare enumerated values more directly (e.g., comparing enum.to_a to [1, 2]) so the test intent is clearer.

[Earlopain]

Ok, I should have applied all your suggestions (also in #1370).