Files
LithosAnanake/proof/StarForth_Dictionary_Manipulation_Words.thy
T
Robert Allan JamesandClaude Sonnet 5 dfdabcc2d7 word_source: repair DECAY-RATE@ overflow guard and remove dead shadowed registrations
DECAY-RATE@ (physics_freeze_words.c) pushed to the data stack with no
capacity check and no prior pop to make room, unlike its neighbors in
the same file -- the one live, unconditional missing-guard bug the
Isabelle sweep's ~15 candidate findings reduced to once checked against
vm_push()'s real internal bounds check (see proof/FINDINGS.md SS2).

Removed dictionary_manipulation_words.c's [ ] STATE and defining_words.c's
DEFER IS DEFER@ (plus the now-orphaned defining_runtime_defer helper) --
all confirmed permanently shadowed by later dictionary registrations
(defining_words.c and defer_words.c respectively), per FORTH's
newest-first lookup. No behavior change: the removed code was already
unreachable.

Verified: hosted `make` builds clean under -Wall -Werror; the hosted
self-test suite passes 965/965 implemented tests with no regressions.
Three-architecture QEMU acceptance boot, all clean to ok> with an
identical dict_hash=0x24b4279f0670aa3a across amd64/aarch64/riscv64 and
identical 1003/965/0/0 test totals -- logs attached.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
2026-08-14 21:33:17 -04:00

247 lines
13 KiB
Plaintext

theory StarForth_Dictionary_Manipulation_Words
imports StarForth_Base
begin
(* =========================================================================
POST-07: Dictionary Manipulation Words
Mirrors: src/word_source/dictionary_manipulation_words.c
This file splits sharply into two classes:
1. Mode/flag words ([, ], SMUDGE, HIDDEN, INTERPRET) -- provable against
the existing vm_mode / dictionary / latest_id fields.
2. Raw-pointer dictionary-navigation words (>BODY, >NAME, NAME>, >LINK,
LINK>, CFA, LFA, NFA, PFA, TRAVERSE, FIND, ') -- NOT modelled. Every
one of these casts a cell_t directly to/from a C `DictEntry*` and
walks struct fields (name_len, link, the byte after the header) that
have no counterpart in the abstract `dict_entry` record (StarForth_
Base.thy:293), which is indexed abstractly by word_id, not by a raw
memory address at all. This is the same class of gap already flagged
for control_words.c's vm_ip/return-stack-as-raw-pointers (see that
file's resume-point note) -- formalizing it would mean re-deriving
the C struct layout (header size, name_len byte, link pointer
arithmetic, body alignment) as a second, address-based dictionary
model living alongside the existing word_id-based one. Out of scope
for this pass; flagged per-word below as documented gaps, not lemmas.
Per CLAUDE.md: `dictionary_m_word_find` (registered as "FIND") is the
real, tested, non-standard FIND word ("parses from input stream").
This file only *reads* it for modelling purposes here -- consistent
with "never modify FIND", since nothing in proof/ touches the C
source.
── Genuine findings, not fixed ──────────────────────────────────────────
1. `[`, `]`, `STATE`, and `INTERPRET` all read/write a `static cell_t
state_variable` at FILE SCOPE (line 49) -- NOT `vm->state_var`
(include/vm.h:431, the real per-VM STATE field actively used by
vm.c/vm_core.c/vm_bootstrap.c/capsule_loader.c on both hosted and
kernel builds). This static is a single process-wide variable, not
even per-VM. CORRECTED 2026-08-14 -- these four words split into two
different reachability classes, not one:
- `[`, `]`, `STATE` are permanently SHADOWED (see StarForth_Defining_
Words.thy's duplicate-registration finding) -- dead code, this
static's misuse never executes for any live VM.
- `INTERPRET` is NOT shadowed (defining_words.c registers no word by
that name at all) -- it IS live and reachable, and its body really
does execute `state_variable = 0;` on every call
(dictionary_manipulation_words.c:392) alongside the correct
`vm->mode = MODE_INTERPRET`. This write is functionally inert
(nothing on any live path reads `state_variable`, since the only
word that ever read it -- this file's own `STATE` -- is itself one
of the shadowed three above), but it is live, executing code in a
registered, tested word, not dead code to remove casually per
.claude/CLAUDE.md's "never modify a registered, tested word"
guidance. Reported, not touched.
`vm->state_var` itself is correctly toggled elsewhere in the real
interpreter loop regardless, so the dictionary-space compile/
interpret bookkeeping isn't broken by any of this. Modelled below
using only `vm_mode` (which `[`/`]` DO correctly set) -- the dead
static write has no vm_state counterpart and is simply omitted, not
"fixed".
2. `dictionary_m_word_hidden`'s `#else` fallback branch (taken only if
`WORD_HIDDEN` is undefined -- it is always defined per include/vm.h:
181, so this branch is dead in every build configuration seen) calls
`dictionary_word_smudge`, which does not exist: the real function in
this file is `static void dictionary_m_word_smudge` (note the `_m_`
and `static` -- not externally callable even if the name were right).
Latent compile-time bug, unreachable under current WORD_HIDDEN
always-defined builds. Not modelled (dead branch), not fixed.
======================================================================== *)
(* ── [ ( -- ) : enter interpretation mode ─────────────────────────────── *)
(* C: vm->mode = MODE_INTERPRET; state_variable = 0 (dead static, see above). *)
definition forth_left_bracket :: "vm_state \<Rightarrow> vm_state" where
"forth_left_bracket vm = vm\<lparr>vm_mode := ModeInterpret\<rparr>"
lemma left_bracket_sets_interpret:
"vm_mode (forth_left_bracket vm) = ModeInterpret"
by (simp add: forth_left_bracket_def)
lemma left_bracket_data_stack_unchanged:
"data_stack (forth_left_bracket vm) = data_stack vm"
by (simp add: forth_left_bracket_def)
lemma left_bracket_never_errors:
"vm_error (forth_left_bracket vm) = vm_error vm"
by (simp add: forth_left_bracket_def)
(* ── ] ( -- ) : enter compilation mode ─────────────────────────────────── *)
(* C: vm->mode = MODE_COMPILE; state_variable = -1 (dead static, see above). *)
definition forth_right_bracket :: "vm_state \<Rightarrow> vm_state" where
"forth_right_bracket vm = vm\<lparr>vm_mode := ModeCompile\<rparr>"
lemma right_bracket_sets_compile:
"vm_mode (forth_right_bracket vm) = ModeCompile"
by (simp add: forth_right_bracket_def)
lemma right_bracket_data_stack_unchanged:
"data_stack (forth_right_bracket vm) = data_stack vm"
by (simp add: forth_right_bracket_def)
lemma right_bracket_never_errors:
"vm_error (forth_right_bracket vm) = vm_error vm"
by (simp add: forth_right_bracket_def)
(* ── [ then ] / ] then [ : mode is exactly what was set last ─────────────── *)
lemma bracket_right_then_left:
"vm_mode (forth_left_bracket (forth_right_bracket vm)) = ModeInterpret"
by (simp add: forth_left_bracket_def)
lemma bracket_left_then_right:
"vm_mode (forth_right_bracket (forth_left_bracket vm)) = ModeCompile"
by (simp add: forth_right_bracket_def)
(* ── STATE ( -- addr ) -- NOT MODELLED ────────────────────────────────── *)
lemma state_not_modelled: True
\<comment> \<open>STATE pushes `(cell_t)(uintptr_t)&state_variable` -- the address of a
bare host C static, not a VM address and not `vm->state_addr`/
`vm->state_var` (which DO exist in the abstract model and ARE the real
per-VM state, see the file-level finding above). There is nothing in
vm_state for this word to legitimately push; modelling it would mean
inventing a fake address for a variable outside the VM's own address
space entirely. Left undocumented as a lemma, gap only.\<close>
by simp
(* ── SMUDGE ( -- ) : compile-only guard, empty body ──────────────────────── *)
(* C: error unless vm->mode == MODE_COMPILE; the "existing body that toggles
the smudge bit" is literally absent (comment placeholder only, no code).
So the word's entire effect, when it doesn't error, is: nothing. *)
definition forth_smudge :: "vm_state \<Rightarrow> vm_state" where
"forth_smudge vm = (if vm_mode vm \<noteq> ModeCompile then set_error vm else vm)"
lemma smudge_requires_compile_mode:
assumes "vm_mode vm \<noteq> ModeCompile"
shows "vm_error (forth_smudge vm)"
by (simp add: forth_smudge_def set_error_def assms)
lemma smudge_in_compile_mode_is_noop:
assumes "vm_mode vm = ModeCompile"
shows "forth_smudge vm = vm"
by (simp add: forth_smudge_def assms)
(* ── HIDDEN ( -- ) : set WORD_HIDDEN (0x40) on the latest word ──────────── *)
(* C: error unless vm->mode == MODE_COMPILE; error if vm->latest is NULL;
else e->flags |= WORD_HIDDEN (the #else dead-branch, see file note #2,
is unreachable and not modelled). *)
definition WORD_HIDDEN :: nat where "WORD_HIDDEN = 0x40"
definition forth_hidden :: "vm_state \<Rightarrow> vm_state" where
"forth_hidden vm =
(if vm_mode vm \<noteq> ModeCompile
then set_error vm
else case latest_id vm of
None \<Rightarrow> set_error vm
| Some wid \<Rightarrow>
(case dictionary vm wid of
None \<Rightarrow> set_error vm
| Some e \<Rightarrow>
vm\<lparr>dictionary := (dictionary vm)
(wid := Some (e\<lparr>de_flags := de_flags e OR WORD_HIDDEN\<rparr>))\<rparr>))"
lemma hidden_requires_compile_mode:
assumes "vm_mode vm \<noteq> ModeCompile"
shows "vm_error (forth_hidden vm)"
by (simp add: forth_hidden_def set_error_def assms)
lemma hidden_requires_latest:
assumes "vm_mode vm = ModeCompile"
assumes "latest_id vm = None"
shows "vm_error (forth_hidden vm)"
by (simp add: forth_hidden_def set_error_def assms)
lemma hidden_requires_latest_present_in_dict:
assumes "vm_mode vm = ModeCompile"
assumes "latest_id vm = Some wid"
assumes "dictionary vm wid = None"
shows "vm_error (forth_hidden vm)"
by (simp add: forth_hidden_def set_error_def assms)
lemma hidden_sets_flag:
assumes "vm_mode vm = ModeCompile"
assumes "latest_id vm = Some wid"
assumes "dictionary vm wid = Some e"
shows "\<exists>e'. dictionary (forth_hidden vm) wid = Some e' \<and>
de_flags e' = de_flags e OR WORD_HIDDEN"
using assms by (simp add: forth_hidden_def)
lemma hidden_data_stack_unchanged:
"data_stack (forth_hidden vm) = data_stack vm"
by (auto simp: forth_hidden_def set_error_def split: option.split)
(* ── INTERPRET ( -- ) ──────────────────────────────────────────────────── *)
(* C: vm->mode = MODE_INTERPRET; state_variable = 0 (dead static, see file
note). Comment in the C source admits the word does not itself invoke
the interpreter loop ("This word exists mainly for completeness and mode
setting") -- so its real effect is identical to `[`. *)
definition forth_interpret :: "vm_state \<Rightarrow> vm_state" where
"forth_interpret = forth_left_bracket"
lemma interpret_is_left_bracket:
"forth_interpret vm = forth_left_bracket vm"
by (simp add: forth_interpret_def)
lemma interpret_sets_interpret_mode:
"vm_mode (forth_interpret vm) = ModeInterpret"
by (simp add: forth_interpret_def forth_left_bracket_def)
(* ── Raw-pointer dictionary navigation -- NOT MODELLED ────────────────────
See the file-level note at the top: >BODY, >NAME, NAME>, >LINK, LINK>,
CFA, LFA, NFA, PFA, TRAVERSE, FIND, ' all cast cell_t <-> DictEntry* and
walk struct-layout details (name_len, link, header-relative body offset)
absent from the word_id-indexed abstract dict_entry model. Each gets a
named sentinel lemma so the gap is greppable per word, matching the
SP@/SP! convention in StarForth_Dictionary_Words.thy. *)
lemma to_body_not_modelled: True \<comment> \<open>>BODY: xt -> DictEntry* -> data field addr (struct-relative pointer arithmetic).\<close>
by simp
lemma to_name_not_modelled: True \<comment> \<open>>NAME: xt -> DictEntry* -> &entry->name (struct field address).\<close>
by simp
lemma name_to_not_modelled: True \<comment> \<open>NAME>: linear scan of vm->latest chain by raw name-field pointer identity.\<close>
by simp
lemma to_link_not_modelled: True \<comment> \<open>>LINK (= LFA): addr -> DictEntry* -> &entry->link (struct field address).\<close>
by simp
lemma link_from_not_modelled: True \<comment> \<open>LINK>: dereferences a DictEntry** raw pointer.\<close>
by simp
lemma cfa_not_modelled: True \<comment> \<open>CFA: identity cast addr -> DictEntry* -> xt; no abstract counterpart needed but the cast itself is untyped.\<close>
by simp
lemma nfa_not_modelled: True \<comment> \<open>NFA: same struct-field address as >NAME.\<close>
by simp
lemma pfa_not_modelled: True \<comment> \<open>PFA: get_body_address -- header size + name_len + alignment arithmetic.\<close>
by simp
lemma traverse_not_modelled: True \<comment> \<open>TRAVERSE: raw byte-pointer arithmetic over a length-prefixed name field.\<close>
by simp
lemma find_not_modelled: True \<comment> \<open>FIND: parses input stream (vm_parse_word) + vm_find_word lookup, neither modelled anywhere in this suite yet. Per CLAUDE.md this is the real, tested, non-standard FIND -- read-only reference here, the C is untouched.\<close>
by simp
lemma tick_not_modelled: True \<comment> \<open>' (tick): identical parse+lookup dependency as FIND, plus a not-found error path FIND itself doesn't have.\<close>
by simp
end