Ruby Issue Tracking System

10/29/2026

06:20 PM Ruby Revision 0dde38a1 (git): ZJIT: Support expandarray with a rest array

Build forward rest expansion for exact Arrays using existing HIR instructions. nozomemein (Nozomi Hijikata)

06:20 PM Ruby Revision 1b88c2e4 (git): ZJIT: Use unboxed integers for the expandarray rest loop

Use IntAdd and IntLt to keep loop control in CInt64 and avoid Fixnum boxing and unboxing. nozomemein (Nozomi Hijikata)

06:20 PM Ruby Revision 64ae80dd (git): ZJIT: Clarify expandarray splat handling

nozomemein (Nozomi Hijikata)

09/02/2026

10:10 PM Ruby Revision f791417e (git): ZJIT: Specialize caller splats with monomorphic lengths

Use the profiled splat array length to expand caller splats into fixed
positional arguments for SendDirect. Guard the runtime length and fall
back to the original Send when it changes or ruby2_keywords semantics
apply.
Validate direct s... nozomemein (Nozomi Hijikata)

10:10 PM Ruby Revision ba2a5dc2 (git): ZJIT: Fix type inference panic after caller-splat dispatch

Caller-splat dispatch creates a join result during type_specialize. A
following CFunc inline may use it before infer_types runs again, causing
ZJIT to infer the type of an untyped Param and panic.
Mark the join result as BasicObject and... nozomemein (Nozomi Hijikata)

10:10 PM Ruby Revision c04e133d (git): ZJIT: Build caller-splat direct paths during dispatch

Move receiver guards, argument emission, and SendDirect generation into a
callback invoked by dispatch_caller_splat. This keeps caller-splat CFG
construction together and allows adding one optimized path per profiled
length for polymorph... nozomemein (Nozomi Hijikata)

10:10 PM Ruby Revision 76d42884 (git): ZJIT: Side-exit on caller-splat guard misses

Use guards instead of a dynamic Send fallback and join block for monomorphic caller splats. Recompile on length mismatches so the call site can collect a new profile, and side-exit without recompiling for ruby2_keywords hashes.
Skip cal... nozomemein (Nozomi Hijikata)

10:10 PM Ruby Revision 1d3b0594 (git): ZJIT: Test caller-splat recompilation with a second length

Document that observing a second splat length makes the profile
non-monomorphic, so the next version keeps the dynamic Send instead of
emitting the same length guard. Add an opt test that verifies this before
the final-version policy app... nozomemein (Nozomi Hijikata)

10:10 PM Ruby Revision 11d6a46a (git): ZJIT: Select caller-splat lengths while building HIR

Select the profiled caller-splat length in add_iseq_to_hir and attach it
to each Send before receiver dispatch creates its specialized arms.
type_specialize then uses the selected length to build each SendDirect
path.
This prepares call... nozomemein (Nozomi Hijikata)

07/13/2026

11:15 PM Ruby Revision 657ce7c7 (git): ZJIT: Clarify SendDirect argument setup comments

Reference the VM argument setup functions that SendDirect mirrors for
keyword-to-positional-hash handling, rest parameter packing, and
optional argument entry selection. nozomemein (Nozomi Hijikata)