Why this is being filed after the fact

feature-a-wasm32-module-writer-wiring granted three compiler.pas edits and closed with "Anything beyond those two is a new conversation." That conversation happened and the grant was given, but it produced only a commit message on the wasm branch (05ec4ac00) and no ticket. So compiler.pas carried five distinct edits on the branch while master's record accounted for three.

The coordinator caught it by diffing the branch rather than reading the merge summary, which is what the standing "nothing is pre-approved" ruling exists for. The lane's own reply had listed four shared-file arms and omitted compiler.pas entirely — the summary was wrong, and only the diff said so.

This is the same defect test/wasm/check_tickets.sh was written for, one level up. That check makes an unfiled ticket unrepresentable. It cannot see an unfiled grant: a shared-file edit whose only record is a branch commit is invisible to master, to the ranker, and to whoever reviews the merge — and it reads as covered, because a neighbouring ticket covers the file.

A grant recorded only in the commit that used it is not a record. The commit is the thing being justified; it cannot also be the justification's index.

The edits, verbatim from the branch

{$include ir_codegen_wasm32.inc} beside the other backends.
function IRTopLevelStmt(k: Integer): Boolean; forward; in the shared forward block.

The dispatch arm was explicitly withheld and was a separate ask. Behaviour change from these two: exactly zero — nothing called IREmitMachineCodeWasm32, and --target=wasm32 still errored in the ir_codegen.inc dispatch, unchanged.

Why the forward rather than a reordering

compiler.pas already forwards IRNodeOwnsManagedStr for precisely this structural reason — the cross backends are included before ir_codegen.inc — and says so in a comment beside it. So this uses the mechanism the file already has for this situation instead of adding a second one (normalise-dont-special-case). The argument that carries is "same mechanism, same reason", not "it is small". Note the forward block is also the region that produced this lane's two FPC-seed REDs, so it is not a free place to add a line.

What the include bought, and how it was checked

228 lines written blind became COMPILED. Unlike wasmenc.inc, this file depends on Syms, Procs, IRKind, CurProc and FrameSize, so no standalone harness for it means anything.

It compiled on the first attempt, which was treated as suspicious rather than accepted, and verified three ways: two distinctive strings from the file were present in the binary; a deliberately broken procedure appended to it was caught (pascal26:232: error: undefined variable (this)); and removing that restored the same fixedpoint. The file is genuinely compiled, not silently skipped.

Gate — as run at the time

make compiler/pascal26 → converged after 1 round(s), 8dedbc7a0e07 (was a5ba15590962).
No existing target moved: 4 programs x 6 targets = 20 output hashes identical, both sides regenerated back-to-back rather than reusing an earlier baseline — Track O had found that a pin landing lib/rtl between two hash runs made six targets look changed.
check_phase1.sh and proto/check.sh green under the new binary.

Consequence for the merge

compiler/compiler.pas is a shared-file arm of the wasm branch and belongs in the merge accounting alongside exception_emit.inc (x2) and ir_codegen.inc (x1). Five edits total in the file, three covered by feature-a-wasm32-module-writer-wiring and two by this ticket. "Covered by a prior ticket" and "not a shared-file change" are different claims, and only the first is true of any of them.