Build w6a and w6l from package-main directories and expose the wcc backend through a narrow package API so w6c and wwdump no longer import implementation files. Retarget the remaining load-bearing fixtures and example sources to directory packages; retain the one intentional flat compiler collision as an explicitly composed raw unit.
The TODO(#11) silent-accept of `for (let (a, a) .. xs)` rested on a
stale premise -- resolvewalk has per-block scopes since #53 and IS
the live w6c_ww checker pass. Bindings now install in a per-LOOP
scope (the N_MCASE pattern), so a duplicate name within one pattern
errs "redeclared in same scope" (cstage parity via stamptuplebinds'
scopedefine-nil check) while sequential same-name loops stay legal.
kwtab restores its stated alphabetical order (`import` before `is`,
both stages, kinds swapped in lockstep with names).
Four seams from one review cluster, all landing on the same checker
files; gates ran on the union (rule 11 body).
?-subset: the wwstage walk compared `...spread` ALIAS nodes on both
sides of the error-subset check, falsely rejecting a spread-carried
error return (`(i64 | ...errs)`, cs-accept/ww-reject) — both sides
now flatten like trycountvariants (cstage Tparams are pre-flattened
at type level).
Multi-let/multi-assign: the tuple gates keyed on the RAW rhs kind,
rejecting a NAMED tuple alias (`type pair = (i64,i64)`; cs clean-
reject, ww asserttyped stop) — both stages chase per #99 alias
transparency, wwstage rettupleof peels the alias so cgmassign keeps
the str element's len/cap stores, and the catA massign-alias error
pin re-rules to a run fixture. The N_MLET diagnostic also printed
its operands swapped (elem/declared reversed vs its wording).
Yield: match_yield_type didn't descend into N_SWITCH, so a yield
inside a switch arm typed the match void and dropped the value
(both stages); and a stray yield outside any match arm reached cgen
unchecked — both stages now gate on a match-arm counter (the
c->loops discipline).
cstage rejects `add(1,2,3)` and `add(4)` (too many / not enough
arguments); wwstage ran no count check at this seam, so both built
-- a stray arg pushed silently, a missing one read garbage
(cs-reject/ww-accept build-verdict divergence). Surplus errs when
params exhaust with args left; missing errs on a leftover regular
param (a leftover TK_ELLIPSIS or FFI "..." is a legal zero-arg
variadic tail). fn-VALUE callees still bail at decl==nil -- their
whole typecheck, arity included, stays task #51.
Two composing defects made a two-enum union mis-tag in wwstage
(live cs!=ww: `let e: (color|shape) = shape.BALL` stored tag 0 —
the color arm — while cstage stored 1). type_eq/typeeq had no
TY_ENUM arm, so ANY two enums fell into the primitive default and
compared equal; enums are nominal (harec: an enum IS its alias
type) and now compare by node identity only. Underneath, the
wwstage post-order revisit re-stamped the constant-folded enum
member (an N_INTLIT) as untyped_int, clobbering the enum stamp the
N_DOT fold applied, so the widen matcher fell to its first-variant
fallback -- the #59.9 N_BIN guard now twins on N_INTLIT (cstage
cexpr is single-pass and never clobbered).
Two silent seams in one sweep. The for-range destructure silently
dropped the 9th+ binding in both stages (bind arrays are sized 8);
the cap now hard-stops per the DEFER_MAX/LOOP_MAX discipline (the
shape is unreachable today -- every wide-tuple construction path
already loud-stops). argtaggedwidensz, the drain-side SSoT for
widened call args, still keyed "natural push" on slot-size equality
while pushargsrev flipped to type equality with #55 -- a same-slot
subset was classified widen by push, natural by drain (counts
coincide today; a classification reader desyncs, the #48 shape).
Both sides now share the wsame key; the widen-branch tag miss for a
concrete source joins the task-2 loud-stop family (tagged subset
sources keep t=0 -- their widentag is never read, the scratch store
remaps).
Lead 17 (enum fold-failure prev+1) DISPROVED: enumvalfold gates
every member init in the checker and its fold set is op-for-op
equal to cgen's enumevalmember, so the fallback only runs inside
already-failing compiles.
cgstructlitfill's `...` total read only an N_INTLIT dim node, so a
def-dimensioned [N]T struct field under-filled: elements 1..N-1 kept
the zero-fill, silently, wwstage only (cstage and the tn twin read
the stamped array length). Read tichase(fi.tnode.type_).alen first;
the surface node stays as the fallback for unstamped shapes.
Hare's rule (harec check.c binarithm): % and the bitwise/shift five
are integer-only; + - * / need numeric operands. ww grouped % with
the numeric ops, and compound assigns never op-checked at all, so
`a % b` on floats compiled half-lowered (live cs!=ww divergence),
`a %= 2.0` plain-stored the rhs (op silently dropped, both stages),
and `s += "cd"` garbled str headers. Gate both at the checker, both
stages; the cgen float-compound fallbacks and the three unknown-
compound legacy defaults (deref/global/local) demote to rule-7 hard
stops. 34 compound-on-tagged/str/slice fixtures re-pin from the old
cgen "not wired" stops to the earlier checker diagnostics; 3 new
reject fixtures pin the closed shapes.
The chained value-struct walker's N_CALL arm handled only <=24B
register returns; an sret-class rhs fell to the scalar tail and
stored ONE word (the sret dest pointer) into the leaf -- silent
truncation in BOTH stages, byteid-blind. Point the callee's hidden
RDI at the BP-relative leaf slot (the single-dot #234 arm verbatim);
a ptr-root/global chain joins the #234-tail loud-stop family.
Every section banner dies (103 -> 0) across test/lang, the observer
suites, the C carriers, and the five comment-heavy corpus fixtures;
banner provenance (#N cites, carrier numbers, repair-cluster labels)
folded into headers or adjacent WHY comments. Narration deleted; row
provenance, ref cites, divergence pins, and layout contracts kept
(fwd-ref decl-order guards and bootstrap-gate corpus rationale
restored where the sweep over-cut). Comment-only proven: all 3742
wwbuild workdir .s byte-identical before/after; test-commit and
test-byteid (161 lang + 1399 data, 0 pinned-divergent) green.
Go compiles only _test.go files as tests; discovery now keys on the
_test.ww suffix alone. The line-leading-@test compatibility allowance
(noncanonical filenames admitted as test sources) is removed from both
driver stages and the coordinator. An @test declaration outside a
*_test.ww file is rejected loudly ("@test declaration outside
*_test.ww", wording byte-identical cs/ww) instead of silently running
under compose or silently dropping in a non-T build (#6). Tree audit
found zero real carriers; the two allowance fixtures flip canonical
(dep_test.ww, widget_test.ww). New pins: direnum attest-noncanon
reject row (both-stage stderr parity) and the coordinator
noncanonical_attest_rejected package row.
The bbdd8bed residue, ruled by the Go derivation (ww's for IS the Go
for; the cstage grammar is the language definition and already the
Go-shaped one):
- bare `for { }` accepted (align UP to parse.c:1023, incl. for-else);
- 3-clause cond and post each omissible — `for (init;; post)`,
`for (init; cond)`, `for (init;)` (align UP, parse.c:1092-1097);
- 2-clause `for (cond; post)` REJECTED (align DOWN): the form exists
in neither Go nor cstage (parse.c:1101 eats a stray ';' then
requires ')'), so the old wwstage accept was a parser bug. The
reject wording mirrors cstage's failure mode per stage.
forhdr_{brace,emptymid,partial} run fixtures pin the accepted forms
byte-identically; forhdr_twoclause_reject pins both reject fragments.
Corpus pin 1745/346/21/209/1169/3490.
A module-scope let whose rhs runs code (alloc, call — peeled through
cast/?/! wrappers) emitted no DATAW slot: emit_lets' fold-fail
silently skipped the definition and every reference died at LINK
time with 'undefined reference', the one unacceptable failure mode
(rule 7). Hare's model rejects at check time (ref/harec/src/
check.c:4360 'Unable to evaluate initializer at compile time') and
routes runtime init through @init, which ww does not have — so both
frontends now reject at the declaration with identical wording.
alias_infptr_global flips compile->error as the alloc pin (its
letvartnode N_TPTR-over-N_TSTRUCT coverage lives on in the local
alias_infptr_{nest,slicecap} siblings); callinit_global_reject pins
the call shape. Corpus pin 1741/345/21/209/1166/3482.
pushargsrev's five aistagged gates (N_IDENT #55, N_CALL #21, N_INDEX
#12, N_DOT #22a, deref #35) treated a tagged arg as already-tagged
when its SLOT SIZE matched the param's. A same-slot subset union
((bool|void) into (i64|bool|void), both 16B) then natural-pushed the
narrower box's words carrying SOURCE tags — no re-layout, no
cg_widen_tag_remap twin — so the callee matched the wrong arm
(silent: probes exited 10/90 where cstage exits 30/27). cstage keys
widen detection on type equality (cgen.c:10000 same = (pu == au) ||
type_eq) and routes every non-same tagged source through the zeroed
scratch + tag remap; the slot-DIFFER wwstage path already mirrored
that byte-identically, so the fix computes cstage's same check once
(wsame) and replaces each slot-size test with it. This also erases
the last known cs!=ww shape divergence (the 16B-local staging vs
direct-push frame delta on prefix subsets).
8 subsetwiden_* fixtures own the class: call-result/ident/str-payload
/mid-arg remap (the wrong-arm shapes), prefix (the shape-divergence
repro), and bigslot/return-pos/struct-24-to-32 sibling guards. Corpus
pin 1740/343/22/209/1166/3480.
`case T1 | T2 =>` (binding-less; cstage parse.c:635-650 chains extra
types through cs->list, bindings stay single-type by design). Three
wwstage layers, no AST change (node.list already exists):
- parser (expr.ww): pipe loop after the non-let arm's first
parsetype;
- checker (check.ww): resolvewalk N_MCASE now walks n.list so each
alt gets its type_ stamp (exhaustiveness/casecovers were already
alt-aware, built ahead of the parser);
- cgen (cgenexpr.ww): the single-pattern want-computation moves
VERBATIM into matcharmwant; a cs.list arm emits CMPQ/JE per alt
funneling into one match_body label, mirroring cgen.c:11118-11141
incl. label mint order and the tag<0 clamp. Nullable arms keep
ignoring alts (both stages).
Graduates the held e2e row (pin 1732/1158/3464); byte-identical on
the repro and the single-pattern control.
cstage parse.c:1027 accepts `for ()` — N_FOR with a nil cond, no
for-else arm, no test emitted (an unconditional loop; a deliberate
Go-flavored ww divergence from Hare, original to the toolchain
import). The wwstage parser had no RPAREN early-out and rejected it.
Selfhost checker and cgen already nil-guard n.cond, so the parser arm
is the whole gap. Graduates the held e2e row (pin 1731/1157/3462).
Adjacent asymmetries left filed, not fixed here (rule 11): the brace
form `for { }`, the empty middle clause of the 3-clause form, and
the REVERSE case (wwstage accepts 2-clause `for (cond; post)` which
cstage rejects).
wwstage dotchainresolve gated its *T roots on an N_TPTR-over-N_TNAME
tnode; an inferred `let p = alloc(S{...})!` rides the checker-
SYNTHESIZED N_TPTR whose pointee is the struct BODY node (the #24
TNAME-normalize covers only the direct struct-lit binding), so the
root gate failed and chained p.field.pseudo reads fell to the
unfused deref arms — shape-only divergence vs cstage's type-keyed
fold (both stages runtime-correct; the historical field(SB) leak the
pin described was already fixed). Widen the local and global root
gates to accept the N_TSTRUCT pointee; correctness stays enforced by
the stamped-tinfo peel below (TY_PTR -> pointee TY_STRUCT).
Graduates alias_g73_heapfill out of DATABYTEID_DIVERGED (2 pins
remain, both the tagged-spill family); adds compile fixtures pinning
the newly-converged siblings (nested read+store, slice .cap,
inferred-global root; pin 1730/22/3460) and a runtime lang row that
mutates then re-points the heap base — a read bypassing the pointer
returns the wrong len.
The last bug-pinned C carrier. Its header documents both #55 halves;
with the cgen side fixed the STRONGER assertion set it deferred is
expressible: barevalue builds the 794 two-file program FLAT
(file-keyed import — a dir-keyed aa/ compiles aa as its own sep unit
where main.v is invisible and the collision cannot express), runs
exit 7 on BOTH drivers, and asserts the flat __root.s cs==ww
byte-identical (rule 10). defshadow pins the def-vs-fn arm-order
corner (foreign scalar def under a curmod fn: bare read 42, not
141), modqualfnval the qualified fn-value positive twin.
Carrier fleet 16 -> 15: 5 units / 6 bootstrap / 3 platform /
738_module_decl. Zero bug-pins remain.
The #173 remap loop in cgtryprop gated on the tparam's explicit `!`
flag only; a LEGACY union (no marks anywhere) classifies error
variants POSITIONALLY (index 0 is success, cstage cg_variant_is_error
cgen.c:905) — so wwstage emitted no remap and propagated the callee's
raw tag into a differently-ordered caller union, a silent wrong arm
(cstage exit 7, wwstage 9 on the banked r700 row). Add the
variantiserror twin and gate the loop on it; refresh the drifted
cstage cite.
Graduates the held e2e row into the corpus (pin 1727/1156/3454) and
adds legacy-mode rows to test/lang/tryprop_tag_remap_test.ww:
reversed order both directions, slot-0 remap, str payload integrity,
a two-hop chain whose pre-fix runtime pass was double-miss tag
cancellation (the lang byteid leg pins the emitted remap blocks),
and a same-order zero-emission control.
test/libenv/libenv_test.ww owns the env-arranged legs of the lib/os
getenv and lib/dirs XDG suites (rows asserted ok, not SKIP, plus the
created-chain layout), the dirs too-long abort on both stages, and the
make-owned runner for the self-arranged stat suite. Wired beside the
sep observers under test-compiler.
The three must-abort runs (u32n/u64n zero, decodedsize unaligned) move
into the owning lib suites as test.expectabort rows; the five
normal-return controls already live in the r989_libprecond_* fixtures.
wwstage parity rides the lib-wide cs==ww byte identity gate.
Each row builds regfile/symlink/subdir under temp.dir (Hare's os-test
idiom) and asserts exact cleanup, so the suite passes bare and the
carrier's env+tree arrangement job disappears. Every carrier assertion
maps to an in-test row.
Strengthens the pre-symmetry-era cstage-only stdout scrape to both
frontends with gated exits and the rule-10 cs==ww byte-id; the libc
binding smoke likewise runs both stages.
One commit for both carriers: the asserttyped manifest and the -c/-r
parse-error gate are the same wwstage-only wwdump_ww subprocess
surface. The 901_*.ww raw gap companions stay in test/wcc, fed via
testenv.repo().
One commit for both carriers: flag-misuse parity and the
missing-package fatal are the same driver-CLI subprocess surface. The
-V oracle parses WW_VERSION out of cmd/wcc/ww.h at run time; the
missingpkg ww_ww twin stays owned by bootstrap carrier 993_ww_ww.c.
Ports 708_param_shadow_mod, 712_redecl and 961_opaque_guards into
test/tool/rejects_test.ww on testenv, one commit because the three
carriers share the single asymmetric-polarity claim (cstage rejects,
wwstage accepts) that a both-stage //ww:error fixture cannot carry.
The neg_selfimp leg drops: test/wcc/data/r948_selfimport/case.ww owns
the identical both-frontend claim.
The unowned remainder is the four run-stderr assertions — msg content
for fail_msg/abort_msg, MANDATORY EMPTINESS for fail_nomsg/abort_bare
(rt_abort(NULL,0) writes nothing) — which the fixture protocol cannot
see (run exit only). The rows build the r957_* fixture sources in
place via the cstage driver; byte-id extends the lowering to wwstage
(the carrier's own argument). Owned legs cited, not duplicated: all
eight exit rows by the r957_assert_*/r957_abort_* fixtures, the
checker rejects by r957_assert_reject_*, byte-id by the
test-data-byteid blanket.
Only the ast_proof leg was unowned: wwdump -a / wwdump_ww -a dump
byte-id plus the ordered dot-over-structlit-over-tname node shape for
a trailing op on a qualified struct literal (inexpressible as a
runtime scenario while the #77 literal-temporary cgen gap holds; the
multilevel row pins the source-order a.b.C tname flatten). The three
build/run scenarios are owned by the r84_qualstruct_* fixtures plus
the test-data-byteid blanket — their vestigial want_exit fields were
already dead in the carrier.
954_tuprecv_run, 956_tuprecv_f64_run and 956_modqualdestr_run shared
one retained dimension: w6c_ww compiles the destructure shape cleanly
AND emits no asserttyped stderr diagnostic (#121 A-narrow / #6a-A
stamp nets) — a channel neither test/lang nor the fixture protocol
has. All 11 rows fold into one observer with the drew-C4 non-vacuity
self-check on the shared predicate. Owned legs cited, not duplicated:
value rows + byte-id ride test/lang/tuprecv_test.ww,
test/lang/tuprecv_f64_test.ww and the test-lang-byteid blanket; the
modqual values + byte-id ride the r956_modqual_* fixtures and
test-data-byteid — those four rows compile the fixture files in place
so probe and fixture cannot drift. Bundled: one file, one concern
(the wwstage stamp-absence dimension).
826_alias_tuple_coerce, 944_alias_accept_run and
944_variant_chain_b95_run all reduced to the same irreducible shape: a
LIVE per-stage asymmetry (probed at HEAD) no symmetric fixture or
@test can host. Ported under the DIVERGE discipline — both sides
asserted loud (826 neg_0: cstage "not assignable to declared pair"
vs wwstage over-accept; #96: cstage runs vs w6c_ww "unsupported
address-of shape"; #277: cstage runs vs ww_ww "deferred #277"), so
a fix on either side fails the row and demands graduation. Owned legs
cited, not duplicated: r826_alias_tuple_* byte rows ride
test-data-byteid, 826's neg_1 went symmetric at HEAD and moves to a
//ww:error fixture per the audit split, the 944 value/reject corpora
live in test/lang/{alias_accept,variant_chain_b95}_test.ww and the
alias_*/vchain_* fixtures. Bundled: one file, one concern (held
divergence pins).
Both carriers were already slim: the r76_def_widen_* byte-id legs are
owned by the test-data-byteid blanket, 783's positive byte rows by the
r78_amp_fn_assign_* fixtures plus that same blanket, and its symmetric
non-tagged rejects by the corpus //ww:error fixtures. The unowned
remainder is three cstage-only loud-rejects (the 760 narrowing def-ref
and 783's two tagged-slot rows); the wwstage over-accepts (no def-init
assignability check; #214 void-variant leniency) are documented at the
site per the residual audit, never asserted as desired — the rows
graduate to //ww:error fixtures when the wwstage side tightens.
Bundled: one file, one concern (cstage-reject pins awaiting wwstage
graduation).
wwiroundtrip pins the @packed re-emit through a real -I tree on both
driver stages (importer-side layout exit 91), strengthened with the
`struct @packed {` needle in the retained sepwork pk2.wwi (a
single-file two-package form emits no .wwi, so the tree is
irreducible). identityreject pins the stage asymmetry as a DIVERGE
row: cstage rejects with "not assignable" (packed is type identity,
harec types.c:621), wwstage builds+runs through the documented
#224/#10 nominal-lossy leniency; a wwstage tighten fails the row and
demands graduation to a //ww:error fixture. The 15 layout/runtime
value rows are fixture-shaped and move to the declarative corpus per
the residual audit port split.
test/misc hosts the residual mixed-shape observers (divergence pins,
stamp/stderr probes, .wwi trees) on testenv, wired into test-compiler
beside the sep/xmod/asm families; its rule declares both full
per-stage toolchains plus the wwdump twins for the coming files.
The 689 port keeps the both-driver run-exit-6 backstop and the
per-stage absolute DATAW row count (linker dedup and byte-id sweeps
are both blind to a symmetric double-emit).
The 18 active rows keep compile + cs==ws byte-id and the tab-anchored
ADDQ $N,SP cleanup markers with both polarities (mem-class rows must
contain their slot's cleanup; the exactly-48B boundary row must not).
The per-row driver build+run legs (incl. the uninit-global zero-box
exit 91) move to the r929 run fixture corpus per the residual audit
port split; the four buildfail rows were already dead here, owned by
the r929_memarg_reject_* fixtures.
The 13 active rows keep compile + cs==ws byte-id and the hidden-RDI
"(BP), DI" presence/absence polarity (present iff sret-class; the two
exactly-32B boundary rows must stay register). The per-row driver
build+run legs move to the r926 run fixture corpus per the residual
audit port split; the seven buildfail rows were already dead here,
owned by the r926_sret_reject_* fixtures.
The five symmetric rows compile their r764_amp_fn_ident_* fixture
sources in place (runtime owned by those fixtures; pre-fix junk-store
still exits 0, so the LEAQ window is the discriminator) and keep the
cs==ws byte-id leg. The cross-module row stays cstage-only pending
#184; its wwstage/byte-id legs graduate with that task.
asmgate_test.ww absorbs 530_w6a_parsenum and 989_datargate_run: the
strtoll-semantics identity rows plus the bare-minus reject against
both assemblers with cross-stage .o byte-identity, and the undefined
DATAR-slot loud reject (no .o written) with the defined-slot
byte-identical ok path. The w6a_ww skip gates drop: the Make target
declares both assemblers.
archive_test.ww absorbs 610_arch and 632_w6l_manyflags: archive
member selectivity through a load-bearing undefined @symbol extern
(link exit 0, run exit 7), and the -L capacity table {1,64,65,100,
128} against both w6l and w6l_ww with an ET_EXEC check per leg. ar is
/usr/bin/ar via runcommand; the junk -L flags are real argv entries
instead of a 16KB shell string.
link_test.ww absorbs 600_w6l and 620_data_link: ET_EXEC on the
trivial pipeline, writable-global read/write runs, the exact
two-PT_LOAD R+X/R+W split, the single-PT_LOAD no-DATAW invariant 992
depends on, and the BSS filesz trim (filesz 1 < memsz 32, loader
zero-fill). Strengthened: the ldd staticness proxy is replaced by a
direct no-PT_INTERP phdr scan.
