Route the 16 routable bare-primsize GUARD sites (is-primitive / struct-vs-prim dispatch) through the #101 aliasprimsize SSoT helper. Byte-NEUTRAL by construction: an alias-narrow name is already neutralized downstream by the same arm, so routing emits no new asm (the empty-flip-set ken oracled). Shape-A exclude-prim-early (3): cgenutil sretretsize / structparamsize / structfloatclass — `primsize>0 return` then structlookup→nil returns the same value; route returns it early, same. Shape-B prim-guard-then-structlookup (13): cgenutil 4604/4650 + cgenexpr 4136/10244 + the 9-site CALL/assign cluster — primsize==0 →structlookup→nil→fall to normal; route skips the block→same normal. Install the peellint bare-primsize FINALE (B7 lint-fuse contract): tools/peellint now rejects any bare primsize() in the ww stage outside the annotated whitelist. Evasion-hardened per the B7 lesson — a character scan (comments + string/char literals stripped first) and a LEFT+RIGHT word-bounded match of the bare `primsize` TOKEN (not just `primsize(`), so the aliasprimsize() wrapper is never a hit and every compiling spelling reds: the call primsize(nm), the paren-wrap (primsize)(nm), the function-value bind `let p = primsize`, and any line-split. ww-only (the C stage dealiases via type_chase_named, no primsize symbol). Two independent exemption windows (peel-ok vs primsize-ok) so neither rule blinds the other. Runs as a make-test dep. Whitelist the 6 designed exemptions with primsize-ok WHY-annotations: machinery — aliasprimsize body (SSoT chase) | typenodeprimresolved + exprprimresolved (#11/#33 prim-resolver chasers) | cgcast leaf-loop + cgenexpr #11 deref-store (own ps==0 fallback; route would regress #11) | the primsize oracle/definition itself (nothing below to chase). structural — elemsizeof x2 + paramfieldsize (chase lives in the -c twin elemsizeofc; threading c is the dormant #110). Empty-flip-set proof: zero C bytes; cstage binaries bit-identical; bootstrap byte-id 990-997 + 950 all green (w6c == w6c_ww on the full selfhost, self-rebuild identical); combined.ww (w6c + wwdump) regen idempotent; sizelint 0; peellint 0 (raw-peel AND bare-primsize over the whole tree = the close-by-construction proof, zero unwhitelisted survivors). Tests: 944_peellint_gate +14 rows (bare / space-before-paren / name-at-EOL split / string-blind opener / paren-wrap / fn-value-bind RED; aliasprimsize wrapper + primsize-ok annotated GREEN; corrupt annotation RED; independent peel/primsize windows; C-file out-of-scope). Closes the #101 primsize-alias family by construction. #109.
This commit is contained in:
154
tools/peellint
154
tools/peellint
@@ -1,38 +1,66 @@
|
||||
#!/bin/sh
|
||||
# tools/peellint — gate against raw TY_NAMED single-peel reads outside
|
||||
# the chase accessors. The #5 alias-arc close-by-construction contract
|
||||
# (rob F2/B7 rulings): one chased accessor is the only spelled way to
|
||||
# dealias — type_chase_named (cmd/wcc/type.c) on the C side, tichase
|
||||
# (selfhost/cmd/wcc/cgenutil.ww) on the ww side. A raw `->under` /
|
||||
# `.under` read peels exactly one NAMED layer; chain-of-aliases stacks
|
||||
# two, so every kind-gated consumer downstream of a single peel falls
|
||||
# to a scalar shape (the four-times-burned family: #60/#61/#62/#70…).
|
||||
# ZERO raw under-token reads may exist in scope outside the annotated
|
||||
# whitelist; this gate lands in the same commit that deletes the last
|
||||
# raw read and keeps the class unwritable.
|
||||
# tools/peellint — gate against two raw single-resolve shapes that
|
||||
# bypass the alias-chase accessors.
|
||||
#
|
||||
# What this does NOT close (stated honestly, per the f2-ruling): a
|
||||
# RULE 1 (raw TY_NAMED single-peel reads) — the #5 alias-arc
|
||||
# close-by-construction contract (rob F2/B7 rulings): one chased
|
||||
# accessor is the only spelled way to dealias — type_chase_named
|
||||
# (cmd/wcc/type.c) on the C side, tichase (selfhost/cmd/wcc/cgenutil.ww)
|
||||
# on the ww side. A raw `->under` / `.under` read peels exactly one
|
||||
# NAMED layer; chain-of-aliases stacks two, so every kind-gated consumer
|
||||
# downstream of a single peel falls to a scalar shape (the four-times-
|
||||
# burned family: #60/#61/#62/#70…). ZERO raw under-token reads may
|
||||
# exist in scope outside the annotated whitelist.
|
||||
#
|
||||
# RULE 2 (bare primsize() name-keyed width) — the #101/#109 close-by-
|
||||
# construction contract. primsize(name) is the ww-stage primitive-width
|
||||
# table; it is ALIAS-BLIND — a narrow alias (`type my32 = u32`) returns
|
||||
# 0, defaulting strides/widths to 8 (the #101 struct-fill miscompile and
|
||||
# the #109 is-primitive GUARD family). aliasprimsize(c, name) is the
|
||||
# SSoT chase (primsize else aliaslookup-chase, cgenutil.ww). A bare
|
||||
# primsize() outside the chase machinery is the forbidden shape: the
|
||||
# size/guard sites route through aliasprimsize so an alias resolves.
|
||||
# This rule is ww-only — the C side dealiases via type_chase_named and
|
||||
# has no primsize symbol. ZERO bare primsize() calls may exist in the
|
||||
# ww stage outside the annotated whitelist.
|
||||
#
|
||||
# Both rules land in the same commit that deletes the last raw shape and
|
||||
# keep the class unwritable.
|
||||
#
|
||||
# What rule 1 does NOT close (stated honestly, per the f2-ruling): a
|
||||
# consumer that never spells `under` at all — a switch on t->kind that
|
||||
# simply never peels — has no token to see here. That NO-PEEL class is
|
||||
# closed only where classification routes through the internalized
|
||||
# chasing helpers, and contained elsewhere by the acceptance-commit-
|
||||
# carries-tripwires doctrine.
|
||||
# carries-tripwires doctrine. Rule 2 has the symmetric edge: a size
|
||||
# computed by a hardcoded literal instead of primsize is caught by
|
||||
# sizelint (rule 13), not here.
|
||||
#
|
||||
# Matcher: a character scan strips block/line comments and string/char
|
||||
# literals first (a regex pass mis-nests `/*` inside a string — review
|
||||
# probe E6), then the under-token is matched accessor-spelling-wide:
|
||||
# `->under`/`.under` in C (deref-dot `(*t).under` is the same peel),
|
||||
# `.under` in ww, with optional whitespace after the operator and a
|
||||
# line-split continuation (`t->` at EOL, `under` next line) — all the
|
||||
# compiling spellings the B7 review probes evaded with. Right token
|
||||
# bound keeps prose like "io.underread" (check.ww) out.
|
||||
# probe E6), then the tokens are matched accessor-spelling-wide:
|
||||
# rule 1: `->under`/`.under` in C (deref-dot `(*t).under` is the same
|
||||
# peel), `.under` in ww, with optional whitespace after the operator
|
||||
# and a line-split continuation (`t->` at EOL, `under` next line).
|
||||
# rule 2 (ww only): the bare `primsize` token, LEFT+RIGHT word-bounded
|
||||
# so the superstring `aliasprimsize` is NOT a hit. Matching the
|
||||
# standalone token (not just `primsize(`) closes the call form
|
||||
# `primsize(nm)`, the paren-wrap `(primsize)(nm)`, the function-value
|
||||
# bind `let p = primsize`, and every line-split — all of which
|
||||
# compile and reintroduce the alias-blind width (review probes).
|
||||
# Right/left token bounds keep prose like "io.underread" (check.ww) and
|
||||
# "aliasprimsize" out.
|
||||
#
|
||||
# Exemption: a line containing `peel-ok` (or the equivalent landed
|
||||
# spelling `peellint-ok`) exempts itself and the following 9 lines —
|
||||
# wide enough that one annotation atop a short construction/chase body
|
||||
# covers it, narrow enough that a stray peel can't hide behind a
|
||||
# distant annotation. Reasons stay WHY-only (rule 8): construction,
|
||||
# chase body, recursive chase, resolve-state probe, or a cited task.
|
||||
# spelling `peellint-ok`) exempts rule-1 violations on itself and the
|
||||
# following 9 lines; a line containing `primsize-ok` exempts rule-2
|
||||
# violations over the same window. The windows are kept separate so a
|
||||
# rule-1 annotation cannot blind a rule-2 bug and vice versa. Wide
|
||||
# enough that one annotation atop a short construction/chase body covers
|
||||
# it, narrow enough that a stray shape can't hide behind a distant
|
||||
# annotation. Reasons stay WHY-only (rule 8): construction, chase body,
|
||||
# recursive chase, resolve-state probe, structural-by-design sizer, or a
|
||||
# cited task.
|
||||
#
|
||||
# Scope: cmd/wcc cmd/w6c selfhost/cmd/wcc lib/ww (skip *.combined.ww).
|
||||
# lib/ww/typ.ww is in scope deliberately — it is type.c's ww mirror,
|
||||
@@ -62,17 +90,21 @@ BEGIN { nviol = 0 }
|
||||
FNR == 1 {
|
||||
cur_file = FILENAME
|
||||
is_c = (cur_file ~ /\.(c|h)$/)
|
||||
exempt_until = 0
|
||||
und_exempt_until = 0
|
||||
prim_exempt_until = 0
|
||||
in_block = 0
|
||||
pending = 0
|
||||
pending_und = 0
|
||||
}
|
||||
|
||||
# Whitelist annotation: arm the exemption window on the RAW line so an
|
||||
# Whitelist annotations: arm the exemption windows on the RAW line so an
|
||||
# annotation inside a comment still counts. `peellint-ok` is the
|
||||
# already-landed sibling spelling (check.ww construction) — accepted
|
||||
# as-is, history is not re-spelled.
|
||||
# as-is, history is not re-spelled. The two windows are independent.
|
||||
tolower($0) ~ /peel(lint)?-ok([^a-z0-9_]|$)/ {
|
||||
if (FNR + 9 > exempt_until) exempt_until = FNR + 9
|
||||
if (FNR + 9 > und_exempt_until) und_exempt_until = FNR + 9
|
||||
}
|
||||
tolower($0) ~ /primsize-ok([^a-z0-9_]|$)/ {
|
||||
if (FNR + 9 > prim_exempt_until) prim_exempt_until = FNR + 9
|
||||
}
|
||||
|
||||
# Strip comments and string/char literals by character scan: a comment
|
||||
@@ -110,28 +142,52 @@ tolower($0) ~ /peel(lint)?-ok([^a-z0-9_]|$)/ {
|
||||
}
|
||||
|
||||
{
|
||||
if (FNR <= exempt_until) { pending = 0; next }
|
||||
# Both C spellings peel: p->under and (*p).under / v.under.
|
||||
if (is_c)
|
||||
re = "(->|\\.)[ \t]*under([^A-Za-z0-9_]|$)"
|
||||
else
|
||||
re = "\\.[ \t]*under([^A-Za-z0-9_]|$)"
|
||||
hit = (code ~ re)
|
||||
# Line-split continuation: operator at EOL, token opening the
|
||||
# next code line. Comment-only lines keep the pend alive.
|
||||
if (!hit && pending && code ~ /^[ \t]*under([^A-Za-z0-9_]|$)/)
|
||||
hit = 1
|
||||
if (hit) {
|
||||
printf("%s:%d: raw under-token read outside the chase accessor; " \
|
||||
"route via type_chase_named (C) / tichase (ww), or annotate " \
|
||||
"peel-ok: <why>\n", cur_file, FNR)
|
||||
nviol++
|
||||
}
|
||||
if (code !~ /^[ \t]*$/) {
|
||||
und_exempt = (FNR <= und_exempt_until)
|
||||
prim_exempt = (FNR <= prim_exempt_until)
|
||||
blank = (code ~ /^[ \t]*$/)
|
||||
|
||||
# RULE 1 — raw under-token peel. Both C spellings peel:
|
||||
# p->under and (*p).under / v.under.
|
||||
if (!und_exempt) {
|
||||
if (is_c)
|
||||
pending = (code ~ /(->|\.)[ \t]*$/)
|
||||
ure = "(->|\\.)[ \t]*under([^A-Za-z0-9_]|$)"
|
||||
else
|
||||
pending = (code ~ /\.[ \t]*$/)
|
||||
ure = "\\.[ \t]*under([^A-Za-z0-9_]|$)"
|
||||
uhit = (code ~ ure)
|
||||
# Line-split continuation: operator at EOL, token opening the
|
||||
# next code line. Comment-only lines keep the pend alive.
|
||||
if (!uhit && pending_und && code ~ /^[ \t]*under([^A-Za-z0-9_]|$)/)
|
||||
uhit = 1
|
||||
if (uhit) {
|
||||
printf("%s:%d: raw under-token read outside the chase accessor; " \
|
||||
"route via type_chase_named (C) / tichase (ww), or annotate " \
|
||||
"peel-ok: <why>\n", cur_file, FNR)
|
||||
nviol++
|
||||
}
|
||||
}
|
||||
|
||||
# RULE 2 — bare primsize token (ww only). LEFT+RIGHT word bounds
|
||||
# so the superstring aliasprimsize() is never a hit and a longer
|
||||
# identifier with a primsize prefix is not matched. Matching the
|
||||
# token (not `primsize(`) catches the paren-wrap `(primsize)(nm)`,
|
||||
# the function-value bind `let p = primsize`, and every line-split —
|
||||
# all compile and reintroduce the alias-blind width (review probes).
|
||||
if (!is_c && !prim_exempt) {
|
||||
if (code ~ /(^|[^A-Za-z0-9_])primsize([^A-Za-z0-9_]|$)/) {
|
||||
printf("%s:%d: bare primsize outside the chase accessor; " \
|
||||
"route via aliasprimsize (ww), or annotate " \
|
||||
"primsize-ok: <why>\n", cur_file, FNR)
|
||||
nviol++
|
||||
}
|
||||
}
|
||||
|
||||
# Pending updates. An exempt line resets its pend (an annotated
|
||||
# operator/name must not carry into a non-exempt next line); a blank
|
||||
# or comment-only line keeps the pend alive; otherwise re-derive.
|
||||
if (und_exempt) pending_und = 0
|
||||
else if (!blank) {
|
||||
if (is_c) pending_und = (code ~ /(->|\.)[ \t]*$/)
|
||||
else pending_und = (code ~ /\.[ \t]*$/)
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
Reference in New Issue
Block a user