Repository navigation
A yield in a block a receiver widened to poly late lifts reaches its block - #10178
Conversation
…block
```ruby
class Rel
def initialize(src)
@src = src
@records = nil
end
def preloaded(records, loaded)
@records = records if loaded
self
end
def loaded_records
records = @records
if records.nil?
records = @src.dup
@records = records
end
records
end
def map
loaded_records.map { |x| yield x }
end
end
class Box
def target = cache
def cache
if missing_value.nil?
end
end
def parts = Rel.new([1]).preloaded(target, false)
end
def doubled(rel) = rel.map { |w| w * 2 }
p doubled(Rel.new([1, 2, 3]))
```
CRuby prints `[2, 4, 6]`; Spinel refused it at compile time: "`yield` in a
block that becomes a Proc, inside a method spliced into a caller that passes
it a block" (refuse_yield_lost_block, 75f03a5). While an_phase_procs looks
at `map`, `loaded_records` answers an Array, so the block is spliced with
`map` and `map` with its caller. Only an_phase_late_widen settles what
`preloaded` stores from `target`, which has no type until then, and turns
`loaded_records` poly. A poly receiver's dispatch materializes the block as a
proc (a_block_is_lifted), where the caller's block is out of reach. In the
roundhouse tiny-api tree the same shape widens in an_phase_method_backstops
(a method reached through `method(:sym)` pinned there).
The loop in an_phase_procs that lowers a yielding method into a function
taking its block as a proc is now an_lower_yielders, and it runs again after
an_phase_nullable_returns, once the late phases have widened what they widen.
A method it did not lower the first time and does now goes through the same
lowering, mark_proc_captures included.
test/yield_in_block_lifted_by_late_poly.rb covers the late-widened receiver,
an empty one, and the sibling `each` that stays spliced.
|
No actionable comments were generated in the recent review.
 |
| Layer / File(s) | Summary |
|---|---|
Extract and repeat yield-method lowering src/analyze.c |
an_phase_procs delegates yield-method lowering to an_lower_yielders. The helper reports whether it lowered methods. analyze_program calls it again after late widening and nullable-return analysis. |
Late-widening regression scenario test/yield_in_block_lifted_by_late_poly.rb, test/yield_in_block_lifted_by_late_poly.rb.expected |
The new test exercises relation mapping, block forwarding, and array iteration. Expected output includes [2, 4, 6], [], and [4, 5]. |
Priority:
Estimated code review effort: 2 (Simple) | ~10 minutes
Change: Bug fix
Merge Risk: 
89e7a
-
The late-widening change appears ready to merge after normal checks; no actionable risk remains identified.
Pre-merge checks: 
-
Check name Status Explanation Resolution Docstring Coverage Docstring coverage is 11.11% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 9 functions across 1 files. (2 skipped: 1… Write docstrings for the functions missing them to satisfy the coverage threshold. -
Check name Status Explanation Description Check Check skipped - CodeRabbit’s high-level summary is enabled. Title check The title names the specific bug fixed: a yield inside a block whose receiver becomes polymorphic only after late widening, where the block is lifted into a proc and must still reach its block. The ra… Linked Issues check Check skipped because no linked issues were found for this pull request. Out of Scope Changes check Check skipped because no linked issues were found for this pull request. Full details: Title check
-
Explanation
The title names the specific bug fixed: a yield inside a block whose receiver becomes polymorphic only after late widening, where the block is lifted into a proc and must still reach its block. The raw summary confirms the change moves yielding-method lowering to run again after late widening and nullable-return analysis, and adds a test for this exact scenario. The title is concise, single-sentence, and specific to the primary change, so it passes. It is somewhat long (74 characters, 15 words) and dense, but it is still clear to a teammate scanning history.
-
Full details: Docstring Coverage
-
Explanation
Docstring coverage is 11.11% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 9 functions across 1 files. (2 skipped: 1 unsupported, 1 too large.)
Resolution
Write docstrings for the functions missing them to satisfy the coverage threshold.
-
- Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts
Warning
Some tools did not complete. Review the errors below.
🔧 Cppcheck (2.22.0)
-
src/analyze.c
-
Cppcheck timed out; analysis of this file is incomplete
-
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.
Comment @coderabbitai help to get the list of available commands.
CRuby prints
[2, 4, 6]; Spinel refused it at compile time: "yieldin ablock that becomes a Proc, inside a method spliced into a caller that passes
it a block" (refuse_yield_lost_block, 75f03a5). While an_phase_procs looks
at
map,loaded_recordsanswers an Array, so the block is spliced withmapandmapwith its caller. Only an_phase_late_widen settles whatpreloadedstores fromtarget, which has no type until then, and turnsloaded_recordspoly. A poly receiver's dispatch materializes the block as aproc (a_block_is_lifted), where the caller's block is out of reach. In the
roundhouse tiny-api tree the same shape widens in an_phase_method_backstops
(a method reached through
method(:sym)pinned there).The loop in an_phase_procs that lowers a yielding method into a function
taking its block as a proc is now an_lower_yielders, and it runs again after
an_phase_nullable_returns, once the late phases have widened what they widen.
A method it did not lower the first time and does now goes through the same
lowering, mark_proc_captures included.
test/yield_in_block_lifted_by_late_poly.rb covers the late-widened receiver,
an empty one, and the sibling
eachthat stays spliced.Verification
make gatenot run. Targeted: everytest/*.rbwhose name has yield, block, proc, lambda, enum, iter or lower (promote_*excluded), throughmake test-corpus TESTS=...: 1661 pass, 0 fail, on this branch rebased onto 2814693..expectedmatches CRuby 3.4 and 4.0.1.tests/tiny_api.rsspinel_gate_models_run, which hit this refusal atruntime/active_record/relation.rb:1581since 75f03a5, compiles and runs with this branch.Summary by CodeRabbit