Re-priced by the owner, 2026-08-30: WASM IS LOW PRIO FROM NOW ON. "it works, it tests our IR, we should be able to compile applications.. for now, that's good enough." The anchor is met — pascal26 runs under wasmtime and emits an ELF byte-identical to the native compiler's for the same source. wasm has served its real purpose, which was exercising the IR from a second direction. These tickets stay OPEN and correct; they simply must not outrank ordinary Track A work. Pick them up on request, or when a lane is warm on the files anyway.

Why this ticket exists

The wasm branch's strategy is move the conflict, don't manage it: it adds new files and touches shared ones only under a named grant. That worked, and the grants were real. What did not work is the accounting.

On 2026-08-28 the lane reported its merge set as four arms and omitted compiler/compiler.pas entirely. The coordinator caught it by diffing the branch instead of reading the summary. The lane's source for that summary was the escapes table in devdocs/dev/wasm/CHARTER.md, which had not been updated since Phase 1 — it still described the exception mechanism as "not yet taken" three phases after it was taken.

A ledger that lives on the branch is not a ledger. It is invisible to master, to tools/progress.sh, and to whoever reviews the merge — the same defect test/wasm/check_tickets.sh was written to make unrepresentable for tickets, recurring one level up for the record of what was granted. See feature-a-wasm32-compile-check-the-phase-2-backend, filed the same day for a grant whose only record was a branch commit message.

So this ticket is the ledger, on master, ranked. It is the merge itself, and it carries the list the merge has to review.

The arms, measured — `git diff origin/master...wasm`

file	edits	what	grant / ticket
`compiler/compiler.pas`	5	3 `{$include}` (`wasmenc`, `asmtext_wasm`, `ir_codegen_wasm32`); a forward for `IRTopLevelStmt` in the shared forward block; the `TARGET_WASM32` output arm `Error(...)` → `writeWasm(outFile)`	3 by `feature-a-wasm32-module-writer-wiring`, 2 by `feature-a-wasm32-compile-check-the-phase-2-backend`
`compiler/exception_emit.inc`	1 arm	the `TARGET_WASM32` arm: `Error(...)` → three poison values (`ExcSetJmpAddr`/`ExcLongJmpAddr`/`ExcRaiseAddr` := -1). Taken twice historically — the message text at Phase 1, the mechanism at Phase 7	message text by `feature-a-wasm32-module-writer-wiring`; the mechanism has no ticket
`compiler/ir_codegen.inc`	1 arm	the `TARGET_WASM32` dispatch arm: `Error(...)` → `IREmitMachineCodeWasm32; Exit;`	no ticket
`compiler/symtab.inc`	1 arm + 1 comment	`EmitZeroFrameSlot`: an explicit `TARGET_WASM32` no-op arm at the top, plus a comment naming the chain's unnamed final arm as x86-64 rather than a default. Emits nothing for any other target	`bug-a-emitzeroframeslot-has-no-wasm32-arm` [A p55]
`compiler/ir_codegen.inc`	+1 arm	`EmitManagedLocalCleanupForTarget`: an explicit `TARGET_WASM32` no-op arm. That chain already fell through to nothing for wasm32, so this changes no bytes on any target and only names what was unnamed	same ticket
`lib/rtl/platform.pas`	1	`PAL_PLATFORM_WASI = 3` beside POSIX=1 and ESP_IDF=2. Additive; no existing line touched	no ticket — and this one is Track B

New files are additive and collision-free and need no arm treatment: compiler/ir_codegen_wasm32.inc, compiler/wasmenc.inc, compiler/asmtext_wasm.inc, lib/rtl/platform/wasi/platform_backend.pas, devdocs/dev/wasm/**, test/wasm/**.

The merge spans TWO lanes' gates

This is the part that is easy to miss, because the branch has been an A-lane effort since Phase 1 and the set was classified uniformly.

compiler/** is Track A: make test + self-host fixedpoint (byte-identical), plus cross where a backend is touched.
lib/rtl/** is Track B: make lib-test / make demos, built against $(PXX_STABLE), never rebuilding the compiler.

lib/rtl/platform.pas and everything under lib/rtl/platform/wasi/ are B's files and carry B's gate. A reviewer running only A's gate has not reviewed them. Track B is parked and does not know this branch exists; the constant alone is not worth waking it for, but a merge is.

What the arms are NOT

Three of the four are the same shape: a TARGET_WASM32 arm that said Error being replaced by the thing it was a placeholder for. The registration skeleton put those Errors there deliberately, because the dispatch chains fall through — a 7th target without an arm gets x86-64 machine code in a file claiming to be its own. Two properties a reviewer should check rather than assume:

In ir_codegen.inc the Exit is now load-bearing rather than implied. An Error never returned; a call does.
In exception_emit.inc the values are -1 and not CodeLen. Nothing can reach them on this target, so the useful failure is a poison that reads as obviously wrong, not a trap. The xtensa arm four lines above is the opposite case and correctly differs.

Standing constraint

Nothing on the branch is pre-approved. The grants above were for the specific edits named in them; "covered by a prior ticket" and "not a shared-file change" are different claims, and only the first is ever true of these. The merge is a deliberate event proposed to the user, never automatic (devdocs/dev/wasm/CHARTER.md, merge policy).

Definition of done