From 38ff4e782ef844a21512eb2ca0e491387ae89ec5 Mon Sep 17 00:00:00 2001 From: _ov4 Date: Tue, 13 Oct 2026 05:25:09 +0000 Subject: fix: compare strings by content Co-Authored-By: Claude Fable 5.1 --- lib/CodeGen/ArithmeticHandler.cc | 2 +- lib/CodeGen/CodeGen.cc | 15 +++++++++++++++ lib/CodeGen/StringHandler.cc | 22 ++++++++++++++++++++++ 3 files changed, 38 insertions(+), 1 deletion(-) (limited to 'lib') diff --git a/lib/CodeGen/ArithmeticHandler.cc b/lib/CodeGen/ArithmeticHandler.cc index 439e4b1..fa71e0c 100644 --- a/lib/CodeGen/ArithmeticHandler.cc +++ b/lib/CodeGen/ArithmeticHandler.cc @@ -83,7 +83,7 @@ Value *ArithmeticHandler::emitBinaryOp(int Op, Value *LHS, Value *RHS, int Line) case tok_ge: return Builder.CreateFCmpOGE(LVal, RVal, "sgetmp"); } } else { - // Same-type pairs (i64/i64, i8/i8 CHAR, i1/i1, ptr/ptr STRING) go + // Same-type pairs (i64/i64, i8/i8 CHAR, i1/i1) go // through untouched; mixed integer widths are widened, anything else // mismatched is a diagnostic instead of invalid IR. Value *L = LHS; diff --git a/lib/CodeGen/CodeGen.cc b/lib/CodeGen/CodeGen.cc index 903d8ef..a6669a6 100644 --- a/lib/CodeGen/CodeGen.cc +++ b/lib/CodeGen/CodeGen.cc @@ -637,6 +637,21 @@ Value *CodeGen::emitExpr(ExprAST *Expr) { if (!L || !R) return nullptr; return StrHandler->emitConcat(L, R); } + // STRING and explicit pointers share LLVM's ptr representation; + // string comparisons must dispatch on the pseudocode type. + switch (Bin->getOp()) { + case tok_eq: case tok_ne: case '<': case '>': case tok_le: case tok_ge: + if ((LT && LT->isString()) || (RT && RT->isString())) { + if (!(LT && LT->isString() && RT && RT->isString())) { + reportError("STRING comparison requires two STRING operands (line %d)", Bin->getLine()); + return nullptr; + } + return StrHandler->emitCompare(Bin->getOp(), L, R); + } + break; + default: + break; + } return ArithHandler->emitBinaryOp(Bin->getOp(), L, R, Bin->getLine()); } diff --git a/lib/CodeGen/StringHandler.cc b/lib/CodeGen/StringHandler.cc index 546506d..0830021 100644 --- a/lib/CodeGen/StringHandler.cc +++ b/lib/CodeGen/StringHandler.cc @@ -1,4 +1,5 @@ #include "cps/StringHandler.h" +#include "cps/Lexer.h" #include "llvm/IR/Type.h" #include "llvm/IR/DerivedTypes.h" #include "llvm/IR/Constants.h" @@ -27,6 +28,9 @@ void StringHandler::setupExternalFunctions() { false); StrCpyFunc = Module.getOrInsertFunction("strcpy", StrCpyType); StrCatFunc = Module.getOrInsertFunction("strcat", StrCpyType); + StrCmpFunc = Module.getOrInsertFunction("strcmp", + FunctionType::get(Type::getInt32Ty(Context), + {PointerType::getUnqual(Context), PointerType::getUnqual(Context)}, false)); FunctionType *CharCaseType = FunctionType::get(Type::getInt32Ty(Context), {Type::getInt32Ty(Context)}, false); ToUpperFunc = Module.getOrInsertFunction("toupper", CharCaseType); @@ -58,6 +62,24 @@ Value *StringHandler::emitConcat(Value *LHS, Value *RHS) { return NewStr; } +Value *StringHandler::emitCompare(int Op, Value *LHS, Value *RHS) { + if (!LHS || !RHS) return nullptr; + + CmpInst::Predicate Pred; + switch (Op) { + case tok_eq: Pred = CmpInst::ICMP_EQ; break; + case tok_ne: Pred = CmpInst::ICMP_NE; break; + case '<': Pred = CmpInst::ICMP_SLT; break; + case '>': Pred = CmpInst::ICMP_SGT; break; + case tok_le: Pred = CmpInst::ICMP_SLE; break; + case tok_ge: Pred = CmpInst::ICMP_SGE; break; + default: return nullptr; + } + + Value *Cmp = Builder.CreateCall(StrCmpFunc, {LHS, RHS}, "strcmp_result"); + return Builder.CreateICmp(Pred, Cmp, ConstantInt::get(Type::getInt32Ty(Context), 0), "str_compare"); +} + // malloc(Len + 1), copy Len bytes from Src + StartIdx, NUL-terminate. // Clamping is each caller's business — MID/RIGHT/LEFT clamp different things. Value *StringHandler::allocCopySubstring(Value *Src, Value *StartIdx, Value *Len) { -- cgit v1.3.1-4-g156e