aboutsummaryrefslogtreecommitdiff
path: root/lib/CodeGen/CodeGen.cc
AgeCommit message (Collapse)Author
2026-10-22feat(input)!: validate INPUT and read whole lines into STRING_ov4
Scalar INPUT now reads one whitespace-delimited token through module-local IR helpers instead of scanf, then skips spaces and tabs and consumes one newline. INTEGER, REAL, BOOLEAN (TRUE or FALSE in any case) and CHAR are validated, and bad input or end of input is a runtime error. STRING INPUT reads the rest of the line. The unused scanf and REAL printf format globals are removed. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-10-22fix(codegen): raise a runtime error when a FUNCTION ends without RETURN_ov4
Report the function and its header line, then throw, instead of silently returning zero. FUNCTION MAIN keeps its implicit result. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-10-22fix(codegen): skip IR verification after reported errors_ov4
Verify the module only when no error was reported, and let defineRecord skip bad fields and still set the record body, marking it Invalid so later missing-field lookups stay silent. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-10-22refactor(codegen): move call marshalling and resolution into CallGen.cc_ov4
Pure move of both marshalCallArgs overloads, emitVariadicArgument and resolveCall out of CodeGen.cc; no behavior change. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-10-22refactor(codegen): remove dead parameters, fields and branches_ov4
Drop the never-read StackObject::IsCopy flag and registerStackObject parameter, the unused NewExprAST forward declaration, the unused Ret parameter of FunctionGen::emitReturn and CG parameter of ArrayHandler::loadElement, unused <cstdio> includes, and the coerceValueToType early returns already covered by the exact LLVM type check (builtin kinds always map to double/i64/i1/i8). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-21feat(driver): color diagnostics like gcc with -fdiagnostics-color_ov4
Add -fdiagnostics-color[=auto|always|never] and -fno-diagnostics-color. The default auto mode colors only a terminal stderr (TERM set and not dumb on Unix; a console that accepts VT sequences on Windows), so test harnesses, the desktop bundle and other pipes stay plain. Every diagnostic label now comes from cps::diagnosticPrefix, which emits gcc's default GCC_COLORS bytes: bold locus, bold red error/fatal error, bold magenta warning, bold cyan note. compilation terminated., runtime messages of the compiled program and lld output are unchanged. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-21feat(diagnostics): print source diagnostics in gcc file:line: layout_ov4
Lexer, parser, driver and codegen diagnostics now read "<path>:<line>: error: <message>" instead of "error: <message> at line N in <path>". The path is always printed as given on the command line. Call sites pass the bare message to cps::sourceError/sourceWarning, which carry printf format checking; reportError routes through the same helper. The suite iii harness parses the new layout and matches the reason against the message only. iii/template-class-pack-field passed only because its reason regex matched the case path, so it is now case-insensitive. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-21feat(codegen): instantiate class templates_ov4
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-21feat(codegen): add variadic function templates_ov4
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-21feat(codegen): support explicit template arguments_ov4
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-21feat(codegen): instantiate function templates by deduction_ov4
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-21feat(parser): parse TEMPLATE declarations, template ids and parameter packs_ov4
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-20refactor(strings): replace compiler builtins with the STRING library_ov4
Construct literals as STRING values and route string operations through ordinary library functions and operators. Preserve value ownership in records, arrays, constants and file I/O. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-18feat(classes): add copy constructors and statement-scoped temporaries_ov4
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-18feat(driver): add #LIB directive hiding library lines in runtime errors_ov4
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-18feat(operators): add member and unary overloads with explicit calls_ov4
Select member and global candidates together, enforce private access, and reuse argument marshalling. Materialize temporary receivers for chaining and register them for the existing object cleanup mechanism. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-17feat(codegen): allow identical EXTERN redeclarations across files_ov4
Repeated EXTERN declarations with the same signature share one C symbol, as C declarations do; conflicting signatures are errors. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-17feat(codegen): declare and call EXTERN C functions_ov4
Variadic arguments must be C values and receive C default argument promotions; narrow integer parameters carry sign/zero extension. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-17feat(classes): overload methods and NEW constructors_ov4
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-17feat(codegen): overload global functions and procedures_ov4
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-17feat(runtime): write runtime diagnostics to stderr_ov4
Runtime check messages and the unhandled-exception message were printed to stdout by the generated program, so `cpsc x.cps -r >out.txt` captured them. They are now formatted with snprintf and written with write(2). Test expectations and docs are updated accordingly. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-10-17feat(codegen): FUNCTION MAIN entry point_ov4
Invoke the internal entry function through the generated exception handler and truncate its INTEGER result to the process exit status. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-17feat(codegen): GLOBAL variables as LLVM globals_ov4
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-17feat(driver): compile multiple source files with INCLUDE visibility_ov4
Reject builds without an entry program and duplicate cross-file definitions. Preserve source order for each unit's own constants and import visible exports. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-17refactor(diag): file-qualified diagnostic locations_ov4
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-17feat(classes): destroy stack objects during exception unwinding_ov4
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-17feat(classes): destroy stack objects at scope exit_ov4
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-17feat(classes): allocate objects with NEW and rebind named class variables_ov4
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-17feat(classes): support ordinary member methods and this references_ov4
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-17feat(memory): add tracked MALLOC and FREE builtins_ov4
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-16fix(compiler): correct CIE semantics and diagnostic locations_ov4
Use standard WHILE headers and default-public class fields, resolve pseudocode identifiers without case sensitivity, and check signed i64 literal boundaries before conversion. Preserve source locations across recursive emission and function diagnostics. Add 22 regression cases and update compiler examples and documentation. Validated all 69 CTest entries with the original Linux Actions seed and verified the release package's embedded, native and IR output paths.
2026-10-16feat(constants): implement literal constants with lexical scope_ov4
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-16refactor(codegen): split emission by responsibility_ov4
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-15fix(codegen): restore runtime diagnostics before throwing_ov4
Reuse historical runtime messages and source-line reporting for arithmetic, array, enum, pointer and file checks while preserving anonymous exceptions. Update the manuals and subsystem guidance to describe diagnostic output. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-14build(toolchain)!: require LLVM and Clang 23_ov4
Reject other compiler and LLVM major versions and remove GCC compiler fallbacks. Adapt string globals, block termination checks and the embedded target triple to LLVM 23 without changing pseudocode behavior. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-14feat(codegen): support C scalar arithmetic and input output_ov4
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-14feat(parser): add explicit parenthesized type casts_ov4
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-14feat(types): add native C scalar types and imported aliases_ov4
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-13feat(compiler)!: add try-catch with C++ stack unwinding_ov4
Add TRY/CATCH/ENDTRY and anonymous THROW statements. Use LLVM exception handling and the C++ ABI to propagate exceptions across function calls, clean up catch state, and handle uncaught exceptions at the program boundary. Register exception runtime symbols for the statically linked embedded runner. Verify runtime checks, nested catches, returns, loops, and invalid syntax with smoke programs; exercise unwinding with MCJIT, lli, and native binaries. BREAKING CHANGE: Runtime check failures now throw anonymous exceptions and print only a generic message when uncaught. Native binaries require linking the C++ runtime. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-13Add record-derived classes with explicit field visibility_ov4
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-13fix: distinguish compiler errors and report compilation termination_ov4
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-13feat: support comma-separated OUTPUT expressions_ov4
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-13fix: check division and modulo for zero divisors_ov4
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-13fix: compare strings by content_ov4
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-13feat: compile source files with configurable IR output_ov4
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-07-26fix: top-level RETURN and READFILE designator targets error cleanly_ov4
- RETURN outside any FUNCTION/PROCEDURE used to emit an i64/void ret inside i32 main — a verifier dump instead of a diagnostic. It is now a clean compile error (no signature means top level: pseudocode cannot name a function 'main', it is a reserved runtime symbol). - READFILE "f.txt", r.f mis-parsed into an error cascade the same way INPUT r.f did before its fix; the target now consumes the designator and reports one clear 'must be a plain STRING variable' error. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-26fix: close type-safety holes found by adversarial review; update docs_ov4
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>
2026-07-26feat: whole arrays (and records) as FUNCTION/PROCEDURE parameters_ov4
- 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>
2026-07-26feat: records with designator lvalues, and pointer types (CIE 4.1/4.2)_ov4
- New CodeGen::emitLValue resolves any designator — variable, checked array element, record field chain (CreateStructGEP via the record payload), pointer deref (with the null-pointer [Fatal] check) — to an address + type name; loadFromLValue centralizes the rvalue read and the STRING null->"" guard, so zero-initialized STRING fields read as "". - Designator assignment stores through emitLValue + emitCoercedExpr; whole-record copies (r2 <- r1, array elements, BYVAL args, RETURN) are aggregate loads/stores with name-checked admission. - getExprTypeInfo walks designator chains statically (field types from the record payload, deref to the pointee), so records/pointers type correctly in every expression context. - Address-of ^x admits into a pointer target only when the designator's type name equals the declared pointee exactly (nominal, strict: the guide's own ^INTEGER-at-a-Season example is a compile error by design); ^arr[i] and ^rec.field work through the same lvalue path. - Verified: the guide's §4.2 StudentRecord example (minus DATE), nested records, arrays of records with Form[i].YearGroup writes, record BYREF/BYVAL params, NodePtr = ^Node forward reference with writes through n1.Next^.Data, unset-pointer deref dying with [Fatal]. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-26feat: user type registry and enumerated types (CIE 4.1/4.2)_ov4
- 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>
generated by cgit v1.3.1-4-g156e (git 2.54.0) at 2026-11-14 21:33:48 +0000