| Age | Commit message (Collapse) | Author |
|---|
|
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
|
Review findings (all reproduced, then fixed and re-verified):
- '^x' leaked as a raw address into any builtin context (s <- ^x printed
the variable's bytes as a C string; NOT ^x, '&', and ^x = ^x compiled).
Address-of is now rejected everywhere except a pointer-typed target.
- A whole-array variable passed the nominal gate as a value of its
ELEMENT type (s <- arrayOfSeason stored the raw data pointer; = on it
emitted invalid IR). Whole arrays are now rejected as operands and as
coercion sources; array parameters are unaffected.
- Enum values smuggled into INTEGER through array subscripts and DECLARE
bounds (a[Autumn], ARRAY[Spring:Winter]); both now go through
emitCoercedExpr like every other INTEGER context.
- An array DECLAREd inside an IF/WHILE/REPEAT/FOR body and used after the
block was a wild-pointer write (constant bounds) or a verifier dump
(runtime bounds). Arrays are now block-scoped: control-flow bodies
snapshot and restore the ArrayTable, so escape is a clean compile error.
- Duplicate parameter names silently shadowed (BYREF write-back defeated);
now a parse error. Trailing comma in parameter lists accepted; now a
parse error consistent with every other list.
- A variable/array/parameter/function could reuse a TYPE name; reserved.
- RETURN with a value inside a PROCEDURE was silently discarded (control
did not return); now an error, as is a bare RETURN in a FUNCTION.
- INPUT r.f mis-parsed into a three-error cascade; now one clear error.
- Assigning to an enum constant now says so instead of 'Unknown variable'.
Docs: ARCH/CONVENTIONS/EXTENDING/SUBSYSTEMS/PITFALLS rewritten for the
TYPE system (ABI additions, nominal-typing invariants, admission-gate
doors, array-parameter ABI, new recipes, new sharp edges); README
features/TODO updated — random files are unblocked.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
|
|
- Parameter syntax lands end-to-end: ARRAY OF T (rank 1, bounds travel
with the argument) and ARRAY[l:u,...] OF T (any rank, literal bounds as
a checked contract). Element type and rank must match exactly.
- Flattened ABI: each rank-R array parameter becomes a data pointer plus
kBoundArgsPerDim i64 bounds per dimension. The callee rebuilds
ArrayMetadata from those bounds (ArrayHandler::bindArrayParameter via a
binder callback injected by CodeGen, so FunctionGen stays ignorant of
the array subsystem); the checked-indexing pipeline is reused unchanged.
- BYVAL (the CIE default) copies the whole buffer at callee entry
(malloc+memcpy, never freed — consistent with the string policy) so
callee writes stay invisible; BYREF binds the caller's buffer.
- Bounded-parameter contract: compile-time comparison when the caller's
bounds are constants, else a runtime [Fatal] check
(RuntimeCheck::emitArrayArgBoundsCheck).
- marshalCallArgs gains the array branch and a call-site line number
(CallStmtAST now carries one); whole-array-to-scalar keeps a clear
error, FUNCTION RETURNS ARRAY is a dedicated parse error.
- Verified: BYREF fill/sort-style mutation, BYVAL copy invisibility,
2-D bounded params, ARRAY OF <record> with field writes through the
parameter, bounds-contract mismatches at compile time and run time.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
|
|
- TypeSystem: TypeKinds Enum/Record/Pointer with per-kind payloads and
checked accessors; two-pass registration (names first, then definitions
in source order) so pointer TYPEs may forward-reference; enum-constant
table owned by TypeSystem so constants stay visible inside function
bodies across the Symbols save/clear/restore. Dead surface removed
(Custom kind, Printable/Builtin/ReferenceLike, ElementSize, hasType,
CodeGen::getNamedValue/getLLVMType).
- Enums are i64 with 0-based ordinals (constexpr kEnumOrdinalBase); all
type identity is by NAME. New emitCoercedExpr is the single admission
gate, replacing raw coercion at assignments, array stores, BYVAL args,
RETURN, IF/WHILE/REPEAT/FOR operands, file operands and builtin args;
coerceValueToType now hard-errors if a user-kind target reaches it.
- Binary ops with user-kind operands are intercepted ahead of
ArithmeticHandler ('&'-intercept precedent): same-enum comparisons,
enum +/- INTEGER (with a runtime range check, [Fatal] on overflow, the
only site that can manufacture an invalid ordinal), same-name pointer
=/<>; everything else is a compile error naming the operator.
- OUTPUT is an exhaustive switch over TypeKind (enum prints its ordinal;
record/pointer are compile errors, also covering whole-array OUTPUT);
INPUT's silent %lld fallback for unknown kinds is now an explicit error.
- Enum value names are globally reserved: collisions with DECLARE,
array names, parameters and function names are compile errors.
- RuntimeCheck: emitEnumRangeCheck + emitNullDerefCheck (used by the
pointer phase) following the emitIndexCheck recipe.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
|
|
- Reserve TYPE/ENDTYPE/SET/DEFINE (-600 token block); SET/DEFINE parse to
dedicated not-implemented errors (RANDOM precedent).
- New TypeAST.h: enum/pointer/record decl nodes, designator chain
(DesignatorExprAST + Index/Field/Deref accessors), AddrOfExprAST,
DesignatorAssignStmtAST. Bare-name and single-index shapes degrade to
their legacy nodes, so existing programs produce identical ASTs.
- ParserType.cc: TYPE declarations (top-level only, enforced via a
ParseBlock depth counter), DECLARE-only record bodies with local
field-level error recovery.
- Postfix [ ] . ^ chain in ParseIdentifierExpr; prefix ^ (address-of,
designator operand only) in ParseUnary; assignment targets rebuilt on
the same suffix parser.
- Parameters: ParamDecl{name, type, mode, array rank/bounds} replaces the
(name, type, byref) tuple. Pass mode is sticky across parameters and
BYREF is rejected in FUNCTIONs, both per guide §8.3 (the old per-param
reset made the guide's own SWAP example pass Y by value). ARRAY OF T /
ARRAY[l:u,...] OF T parameter syntax parses; codegen rejects it until
the parameter-passing phase lands.
- CodeGen stubs report clear not-implemented errors for the new nodes so
nothing is silently ignored.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
|
|
(OPENFILE/READFILE/WRITEFILE/CLOSEFILE/EOF)
Front end: OPENFILE/READFILE/WRITEFILE/CLOSEFILE plus reserved mode
keywords READ/WRITE/APPEND (RANDOM reserved with a dedicated
not-implemented error); statement nodes in new FileAST.h, parsing in new
ParserFile.cc; EOF joins the Builtins table as the 12th builtin and
CallExprAST now carries a line number for its runtime diagnostics.
Runtime: new FileHandler (StringHandler-shaped leaf) lazily synthesizes
internal __cps_file_* helper functions and a 16-slot open-file table
keyed by filename content (strcmp — STRING '=' is pointer identity).
EOF is an fgetc/ungetc peek; READFILE grows a malloc/realloc line buffer
and strips a trailing CR; fopen uses text modes so mingw-w64 line
translation works. File-free programs get byte-identical IR. All misuse
(not open, wrong mode, reopen, fopen failure, read past EOF, table full)
dies with the existing "[Fatal] line N:" pattern via a new
RuntimeCheck::emitErrorAndExit overload taking runtime printf args.
Also fixes a pre-existing ArithmeticHandler hole surfaced by adversarial
review: mixed-width integer operands (i1 BOOLEAN / i8 CHAR vs i64) now
widen via ZExt instead of emitting invalid IR, and remaining type
mismatches get a clean diagnostic; -EOF(f) and -IS_NUM(x) compile.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
|
|
Confirmed by a 17-agent review pass over the refactor diff (12 findings
verified against the code and reproduced end-to-end):
- marshalCallArgs: BYREF arguments must match the declared parameter
type exactly (was: silent bit-reinterpretation, e.g. REAL passed to
BYREF INTEGER printed garbage with exit 0); whole-array arguments are
rejected (was: callee assignment smashed the array data pointer ->
runtime SIGSEGV)
- Null guards for CHR/LCASE/UCASE/IS_NUM/NUM_TO_STR/STR_TO_NUM/ASC
argument emission (was: compiler SIGSEGV on e.g. CHR(undeclared)),
matching the guards MID/RIGHT/LEFT already had
- ParseReturnStmt: propagate expression-parse failure (was: diagnostic
printed but exit 0) and recognize bare RETURN before
ENDWHILE/UNTIL/NEXT (was: bogus diagnostic on legal early return)
- emitPrototype rejects FUNCTION/PROCEDURE names that collide with
runtime/libc symbols (was: FUNCTION strlen silently hijacked the
string runtime; arity mismatch crashed createArgumentAllocas) or with
builtins (was: silently shadowed); conflicting redefinitions keep the
first signature
- FOR STEP is evaluated exactly once, before the loop (was: once per
iteration; the runtime-direction refactor had moved it to N+1)
- RETURN values are coerced to the declared return type via the
signature table (RETURN n in a RETURNS REAL function now compiles)
- '&' coerces its operands to STRING (CHAR/number operands work);
OUTPUT of a PROCEDURE call reports a clean error
AGENTS docs updated to match. All positive samples remain byte-identical
to the pre-refactor baseline.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
|
|
Integer/Real/Boolean/CharHandler carried only five one-line LLVM API
wrappers (now inlined at their call sites in CodeGen.cc) plus dead
emitDeclare methods that violated the documented two-table lock-step
invariant. StringHandler::emitDeclare (also dead) and its NamedValues
reference go too.
With the dead writers gone, NamedValues held no information not already
in Symbols[].Storage, so the parallel map is removed: getNamedValue
reads Symbols, ArrayHandler/FunctionGen take one table, and function
bodies save/restore a single map. Behavior unchanged.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
|
|
emitPrototype now records each FUNCTION/PROCEDURE's pseudocode signature
(FuncSig: return type + per-param type/BYREF flag). CodeGen::
marshalCallArgs replaces the two verbatim-duplicated marshalling blocks
and dispatches on the declared signature instead of inferring BYREF from
pointer-typed LLVM params: BYVAL STRING params now receive the string
data pointer (no more runtime garbage), and BYVAL args are coerced to
the declared type (INTEGER->REAL args no longer emit invalid IR).
Top-level prototypes are pre-registered in compile(), so calls may
precede definitions; GenerateCall no longer fabricates i64 externs for
unknown names - an unregistered callee is a compile error. OUTPUT
reports an error when it cannot infer the expression type instead of
printing with the INTEGER format.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
|
|
- MID/RIGHT/LEFT integer args now coerced at call sites (RIGHT(s, n/2)
no longer emits invalid IR); drop emitLeft's private conversion
- FOR loop direction decided at run time via select on STEP sign;
descending FOR works, STEP evaluated once per iteration
- ArrayHandler: extract emitCheckedIndices/loadElement; index count is
validated before emission, fixing a compiler-crash (vector OOB UB) on
OUTPUT with too many indices
- Error contract: Lexer/Parser/CodeGen track errors; all silent failure
paths now print Error: diagnostics with line numbers; unterminated
char literal no longer truncates the rest of the program; module is
verified with verifyModule; driver exits non-zero on any error
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
|
|
|
|
|
|
|
