GRANT: ir_codegen_riscv32.inc → frankS (Track S), scoped to the SPECIAL_IN arm

Granted 2026-08-30. ir_codegen_riscv32.inc is Track A's file and frankS correctly declined to take it on its own read, offering instead to file an A ticket citing the arm32 model. Granting rather than splitting, and the reason is the finding itself.

Why the grant, not the split

SPECIAL_IN is missing from both 32-bit backends — riscv32 and xtensa — and confirmed failing on the same two programs on each. The ticket that produced it exists because one arm of a double case was fixed and the sibling was not grepped for.

Landing the repair on one backend and filing the other is that same defect, one level up. It leaves riscv32 as "the next lane's surprise" — the exact phrase, and the exact outcome, normalise-dont-special-case.md exists to prevent.

A fix for a double-case bug must not itself be applied to one arm. So the two 32-bit backends land the rule together, in one change, by the lane that has the model in context.

Scope

• compiler/ir_codegen_riscv32.inc — the SPECIAL_IN arm only. Nothing else in the file.
• Port model is arm32 / i386 (the two 32-bit arms that already carry it): walk the arg list, IR_BLOCK item = lo..hi range, plain item = equality, accumulate into a scratch register. No conditional execution on either target, so the moveq becomes a branch.
• Gate = the two named programs green on both backends, make compiler/pascal26 (which is the fixedpoint), plus the xtensa/riscv32 differential rows.

Collision check at grant time

frankA holds A/P and is in the pasparser_* files on the typed-const generic repair. frank-optimize-b4 has released ir_codegen.inc and is on an x86-64 emitter ticket. Nobody is in ir_codegen_riscv32.inc. ir_codegen_xtensa.inc is already frankS's.

Note for whoever reads this next

riscv32's diagnostic for the missing arm is the generic "standard builtin calls not supported in bare-metal stage 1" bucket — face 144 still live on that backend, and the reason the gap was invisible there. Fixing the arm does not fix the message; that is bug-a-iropname-has-no-entry-for-seven-ir-ops territory and stays A's.

Log

• 2026-08-30 — resolved, commit f843a8366.