The wcc test drivers ran `ww build <bare-/tmp src>` with no -o, so the
compiler's <stem>.sepwork scratch landed beside the source and was never
cleaned: unbounded /tmp growth (2195 stale dirs observed) that fills tmpfs
and fabricates phantom test failures + silent harness aborts, and for
in-repo fixture builds leaked .sepwork into the tracked tree.
Each leaking build now writes its source + output inside a per-invocation
tmpdir, passes -o <tmpdir>/<stem> so the .sepwork lands inside it, and
rm -rf's the tmpdir on every exit path -- including fopen-fail and the
expected-fail reject builds (scratch is mkdir'd before the build can fail).
`ww run` and explicit-`-o`/byte-id helpers are left as-is; the 990/993
byte-id comparison logic is byte-for-byte unchanged.
Two items filed separately (this commit holds the no-Makefile / no-main.c
rail):
- #13: a stale <src>.s byte-id readback (749) silently no-ops since
separate-compile emits .s to <ostem>.sepwork/__root.s; documented inline.
- #14: build-system Makefile recipes build selfhost/cmd/*/main.ww with no
-o and leak main.sepwork in-tree (bounded, gitignored; own commit).
One concern -- sepwork leak hygiene -- across 228 drivers; uniform
transform applied per-file and two-round reviewed. make test: all 402
passed, zero net-new /tmp scratch, zero test-driven in-repo .sepwork.
The #13 .cap read-fix gated the wwstage CX→AX shuffle on the base's
type being TY_SLICE/TY_STR, on the assumption that a bare string literal
types as untyped_str and so misses the gate (matching cstage, whose
cap-shuffle lives only in the typed pseudo-field branch). That assumption
is false on wwstage: its checker stamps N_STRLIT as `str` (check.ww:2322),
not untyped_str as cstage does (check.c:1079). So `"abc".cap` passed the
TY_STR gate and emitted a stray `MOVQ CX, AX` on wwstage only — while
cstage's untyped catch-all never shuffles it — a rule-10 byte-id break.
A string literal's cgexpr loads only AX=ptr/BX=len (cgen.ww N_STRLIT),
never a CX cap, so the shuffle was garbage on top of divergent. Exclude
N_STRLIT from the gate: `"abc".cap` now returns AX unshuffled on both
stages, byte-identical. The typed `t[i].cap` path (lhs N_INDEX) is
unaffected.
The underlying N_STRLIT type divergence (cstage untyped_str vs wwstage
str) is a separate latent checker issue, filed for follow-up; this commit
keeps the cgen byte-identical regardless.
683: new BYTEID_ONLY row str_lit_cap_symmetry pins the edge (asm-byte-id
asserted; runtime value is a link-time address). 39/39 ok; test-unit
251/251; smoke cs==ww; sizelint clean. combined.ww (w6c + wwdump) regen'd.
`t[i].cap` (t a `[N][]u8` / `[N]str`) miscompiled in BOTH stages,
divergently — the read-side sibling of #20's store fix. cgexpr on the
indexed element leaves the full {ptr,len,cap} header (AX/BX/CX via
cgslicehdr), but the `.cap` field-selector never shuffled CX→AX:
cstage's typed pseudo-field else-branch handled only .ptr/.len, so
`.cap` fell through returning AX=.ptr; wwstage's cgdot non-ident
catch-all likewise handled only .ptr/.len, emitting no read (stale AX).
`t[i].len` already worked (BX→AX shuffle) — only `.cap` was missing.
Fix mirrors the .len shuffle: add the .cap CX→AX arm in both stages.
The shuffle fires ONLY for a typed slice/str base (TY_SLICE/TY_STR
after NAMED-chase); an untyped str literal (`"abc".cap`) leaves only
AX=ptr/BX=len and must return AX unshuffled — keeping the wwstage
catch-all byte-identical with cstage, whose cap-shuffle lives in the
typed branch, not the untyped catch-all.
Validated direct `t[i].cap` (slice + str, elements 0/1) against the
whole-element-copy oracle (`let q=t[i]; q.cap`, made correct by #20),
plus .len-after-index regression pins, in test 683; dual-stage runtime
+ byte-id (36/36 ok). combined.ww regenerated.
A `let t: [N][]u8 = [a, b]` / `[N]str` literal init lowered each
element's {ptr,len,cap} header into AX/BX/CX (cgexpr) but stored only
some words: a slice element fell through to the scalar 1-word MOVQ
(dropping .len AND .cap), a str element stored 2 words (dropping .cap,
latent). Each element is 24B (post-#1) and must be copied whole.
wwstage was worse — a slice element matched no esz branch, so esz
stayed the 8 sentinel: the per-element stride collapsed (element i+1
overwrote element i's tail), the -96-vs-80 cs!=ww frame divergence.
This is the str/slice arm of the #270 aggregate-element-store family.
struct/array/tuple already copy correctly via the #270-1c is_agg
multi-word path; str/slice were the documented follow-up (cgen.c:9037,
cgenstmt.ww deferral). They can't join is_agg (that path word-copies
from a source slot and rejects non-ident/structlit elements, whereas
str/slice elements are commonly exprs cgexpr lowers into registers) —
the correct mechanism is the existing register header store, extended.
Fix (BOTH stages, converged byte-identical): cstage adds
is_slice_el = type_isslice(esub) and stores 3 words (incl CX->base+16,
the cap) for `is_str_el || is_slice_el`, in the main loop and the
repeat-fill. wwstage adds isslicel (esubti.kind == TY_SLICE -> esz =
esubti.size, fixing the stride) and the matching 3-word store. Closes
[N][]u8 (the bug) and the latent [N]str cap-drop in one branch.
The latent str cap-drop is now stored, but the indexed-element `.cap`
READ (`t[i].cap`) stays broken — a distinct cgindex/dot-selector bug,
cs!=ww divergent, filed as task #13. The new test validates the stored
cap via a whole-element copy (`let q = t[i]; q.cap`), which reads
through the correct ident-load path. [N]tagged literal init is the
remaining sibling (is_agg excludes TY_TAGGED), task #12.
Test 683_arr_strslice_elem: table-driven, dual-stage runtime + asm
byte-id; slice/str .len, 3-element stride-24, cap-via-copy, .ptr deref,
plus a [N]struct regression pin proving the is_agg path is untouched.
