checkarrlitfits' nested recursion keyed on the raw elemtn kind; a
named-alias element type ([2]row, row=[2]int) arrives as N_TNAME, so
the inner overlong literal skipped the count+range checks and the
module static-DATA route emitted silently TRUNCATED data (ken's
m7c_global: DATAW 1,2,4,5 — exit-masked once the #60 read fix removed
the segv; cstage loud-rejects every spelling via its typed-literal
assignability net). #105: the W1 fill gate never runs on this route,
severity raised post-#60.
Fix: chase elemtn through resolvealias (transitive) at the recursion
gate — alias spellings of any depth take the same checks as the
direct shape at all four contexts funneling through the choke point
(module let / local let / def / struct-field). A direct N_TARRAY
passes through resolvealias unchanged, so accepted shapes are
byte-identical base→tip (m7c_global_ok + exact-fit alias
field/def/2lvl probed ASM-ID vs a base scratch build). The m7/m7b
local overlong rows stay loud, now via the earlier count-naming
checker diagnostic instead of the cgen #270-1c fatal. The
out-of-range narrow inner element louds "array element out of
range" exactly as the direct spelling already did on wwstage.
808_arrlit_overlong: 37 -> 50 checks (+1 accept control
alias_exact_module = ken's m7c_global_ok with a byte-id cell, +4 loud
flips alias_nested_{module,2lvl,def,field} pinning per-stage texts,
+1 REVIEW AMENDMENT alias_nested_local pinning the m7/m7b text move
— pre-fix ww was loud via the late cgen #270-1c fatal; the row reds
if the diag regresses off the checker count text).
989 ratchet zero flips — no lib module-level literal trips the gate.
Filed sibling, not folded: OUTER alias-of-array overlong
(let g: arr = [5 elems], arr=[4]int) still ww-silent-truncates at the
alias-blind call-site N_TARRAY gates; cs louds with the count text.
An array literal with more elements than the declared [N] passed the
per-element accept-if-fits checks in both stages and cgen then stored
every element at its natural offset, writing past the slot: local
frames smashed silently (the repeat form [1,2,3...] into [2]int wrote
at the saved BP), module DATA corrupted neighbours. All four
declaration contexts (local let, module let, def, struct-field
literal) funnel through one choke point per stage — arrlit_init_fits
(check.c) / checkarrlitfits (check.ww) — which now pre-counts the
literal (skipping the ... marker) and rejects count > N naming both
counts.
cstage clet's blanket has_arr_repeat bypass is narrowed to non-array
declared targets: repeat literals into arrays now run the same
overlong + #130 range checks wwstage's checkletassign always ran
(the bypass let [2]u8 = [999...] dodge the range check cstage-only).
checkarrlitfits also recurses into NESTED array-literal elements
(declared elem node N_TARRAY): cstage catches the nested shape
through its typed-literal assignability net, which wwstage's untyped
elements have no analog of — [2][2]int = [[1,2,3],[4,5]] at module
scope silently emitted corrupted DATA (1,2,4,5) and the struct-field
twin likewise. Recursion through the one choke point closes any
depth; a named-alias element type still bypasses — task #16.
alen==0/nil-length stays exempt ([0]/[_] sentinel conflation and
un-inferred [_] in def/struct-field — task #11); a non-INTLIT length
child (def-named [N]) is exempt in wwstage — task #13; under-long
literals keep their current accept (Hare rejects — task #10);
wwstage's overlong accept at assign/call-arg/return position (cstage
already rejects) is task #12; exact-fit bare-int nested cs-reject/
ww-accept divergence is pre-existing — task #17.
