build: re-enable warnings as errors (#5)

gf712 · web-flow · commit 25ce0411d00a · 2025-12-16T19:01:20.000Z
And address some of the warnings
diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -49,7 +49,7 @@ if (${ENABLE_SANITIZER_UNDEFINED_BEHAVIOR})
 endif()
 
 project_options(
-	# WARNINGS_AS_ERRORS
+	WARNINGS_AS_ERRORS
 	# ENABLE_COVERAGE
 	# ENABLE_CPPCHECK
 	# ENABLE_CLANG_TIDY
@@ -74,7 +74,7 @@ project_options(
 
 # TODO: Address all the warning below. This should be only temporary...
 target_compile_options(project_warnings
-  INTERFACE $<$<COMPILE_LANGUAGE:CXX>:-Wno-sign-conversion;-Wno-shorten-64-to-32;-Wno-shadow;-Wno-implicit-fallthrough;-Wno-implicit-int-conversion;-Wno-old-style-cast;-Wno-gnu-zero-variadic-macro-arguments;-Wno-implicit-int-float-conversion;-Wno-deprecated-copy;-Wno-missing-field-initializers>
+  INTERFACE $<$<COMPILE_LANGUAGE:CXX>:-Wno-sign-conversion;-Wno-shadow;-Wno-implicit-fallthrough;-Wno-old-style-cast;-Wno-deprecated-copy;-Wno-missing-field-initializers;-Wno-null-dereference;-Wno-maybe-uninitialized>
 )
 
 check_cxx_source_compiles(
diff --git a/src/executable/mlir/Target/PythonBytecode/LinearScanRegisterAllocation.hpp b/src/executable/mlir/Target/PythonBytecode/LinearScanRegisterAllocation.hpp
@@ -466,21 +466,21 @@ class LinearScanRegisterAllocation
 			auto current_value = std::get<mlir::Value>(cur.value);
 
 			// Resolve block arguments to their defining operations
-			if (current_value.isa<mlir::BlockArgument>()) {
+			if (mlir::isa<mlir::BlockArgument>(current_value)) {
 				if (auto it = live_interval_analysis.block_input_mappings.find(cur.value);
 					it != live_interval_analysis.block_input_mappings.end()) {
 					for (auto mapped_value : it->second) {
 						ASSERT(!std::holds_alternative<ForwardedOutput>(mapped_value));
 						if (clobbers_r0(std::get<mlir::Value>(mapped_value))) {
-							ASSERT(current_value.isa<mlir::BlockArgument>());
+							ASSERT(mlir::isa<mlir::BlockArgument>(current_value));
 							current_value = std::get<mlir::Value>(mapped_value);
 							break;
 						}
 					}
 				}
 			}
 
-			ASSERT(!current_value.isa<mlir::BlockArgument>());
+			ASSERT(!mlir::isa<mlir::BlockArgument>(current_value));
 			auto loc = current_value.getLoc();
 
 			// Insert: push r{cur_reg}, move r{scratch}, r{cur_reg}, pop r{cur_reg}
diff --git a/src/runtime/PyArgParser.hpp b/src/runtime/PyArgParser.hpp
@@ -8,6 +8,8 @@
 #include "TypeError.hpp"
 #include "vm/VM.hpp"
 
+#include <limits>
+
 namespace py {
 
 template<typename... ArgTypes> struct PyArgsParser
@@ -56,7 +58,15 @@ template<typename... ArgTypes> struct PyArgsParser
 					return Err(type_error("'{}' object cannot be interpreted as an integer",
 						int_obj.unwrap()->type()->name()));
 				} else {
-					std::get<Idx>(result) = as<PyInteger>(int_obj.unwrap())->as_i64();
+					auto value = as<PyInteger>(int_obj.unwrap())->as_i64();
+					static_assert(sizeof(ExpectedType) <= 8);
+					if (!fits_in<ExpectedType>(value)) {
+						return Err(type_error("{} not within range ({}, {})",
+							value,
+							std::numeric_limits<ExpectedType>::min(),
+							std::numeric_limits<ExpectedType>::max()));
+					}
+					std::get<Idx>(result) = static_cast<ExpectedType>(value);
 				}
 			} else {
 				[]<bool flag = false>() {
diff --git a/src/runtime/PyList.cpp b/src/runtime/PyList.cpp
@@ -505,7 +505,7 @@ PyResult<PyObject *> PyList::__mul__(size_t count) const
 	if (count <= 0) { return PyList::create(); }
 	std::vector<Value> values;
 	values.reserve(count * m_elements.size());
-	for (auto _ : std::views::iota(size_t{ 0 }, count)) {
+	for ([[maybe_unused]] auto _ : std::views::iota(size_t{ 0 }, count)) {
 		values.insert(values.end(), m_elements.begin(), m_elements.end());
 	}
 
diff --git a/src/runtime/PyString.cpp b/src/runtime/PyString.cpp
@@ -150,8 +150,11 @@ PyResult<PyString *> PyString::create(const Bytes &bytes, const std::string &enc
 			if (byte < std::byte{ 128 }) {
 				result.push_back(static_cast<char>(byte));
 			} else {
-				result.push_back(0xc2 + (static_cast<unsigned char>(byte) > 0xbf));
-				result.push_back((static_cast<char>(byte) & 0x3f) + 0x80);
+				result.push_back(
+					static_cast<unsigned char>(0xc2) + static_cast<unsigned char>(byte) > 0xbf);
+				result.push_back(static_cast<char>(
+					(static_cast<unsigned char>(byte) & static_cast<unsigned char>(0x3f))
+					+ static_cast<unsigned char>(0x80)));
 			}
 		}
 		return PyString::create(result);
@@ -1864,9 +1867,14 @@ PyResult<PyObject *> PyString::translate(PyObject *table) const
 			if (mapped_cp > 0x110000 || mapped_cp < 0) {
 				return Err(value_error("character mapping must be in range(0x110000)"));
 			}
-			auto el = utf8::utf8chr(mapped_cp.get_ui());
+			auto mapped_cp_ui = mapped_cp.get_ui();
+			if (!fits_in<uint32_t>(mapped_cp_ui)) {
+				return Err(value_error(
+					"character mapping value {} does not fit in uint32_t", mapped_cp_ui));
+			}
+			auto el = utf8::utf8chr(static_cast<uint32_t>(mapped_cp_ui));
 			if (!el.has_value()) {}
-			cache[mapped_cp.get_ui()] = *el;
+			cache[static_cast<uint32_t>(mapped_cp_ui)] = *el;
 			result.append(*el);
 		} else if (mapped_value.unwrap()->type()->issubclass(types::str())) {
 			auto mapped_str = static_cast<const PyString &>(*mapped_value.unwrap()).value();
diff --git a/src/runtime/Value.cpp b/src/runtime/Value.cpp
@@ -455,8 +455,8 @@ bool Bytes::operator==(const PyObject *other) const
 constexpr unsigned char to_digit_value(char value)
 {
 	if (value >= 0 && value <= 9) { return value; }
-	if (value >= 'a' && value <= 'z') { return (value - 'a') + 10; }
-	if (value >= 'A' && value <= 'Z') { return (value - 'A') + 10; }
+	if (value >= 'a' && value <= 'z') { return static_cast<unsigned char>((value - 'a') + 10); }
+	if (value >= 'A' && value <= 'Z') { return static_cast<unsigned char>((value - 'A') + 10); }
 	return 37;
 }
 
@@ -611,12 +611,6 @@ bool NameConstant::operator==(const NameConstant &other) const
 		[](const auto &rhs, const auto &lhs) { return rhs == lhs; }, value, other.value);
 }
 
-std::ostream &py::operator<<(std::ostream &os, const Tuple &tuple)
-{
-	os << tuple.to_string();
-	return os;
-}
-
 std::string Tuple::to_string() const
 {
 	std::ostringstream os;
@@ -642,6 +636,12 @@ std::string Tuple::to_string() const
 
 namespace py {
 
+std::ostream &operator<<(std::ostream &os, const Tuple &tuple)
+{
+	os << tuple.to_string();
+	return os;
+}
+
 PyResult<Value> add(const Value &lhs, const Value &rhs, Interpreter &)
 {
 	return std::visit(
diff --git a/src/runtime/modules/signal/module.cpp b/src/runtime/modules/signal/module.cpp
@@ -11,6 +11,7 @@
 
 #include <csignal>
 #include <cstdint>
+#include <limits>
 #include <variant>
 
 namespace py {
@@ -83,7 +84,10 @@ PyResult<PyObject *> signal(PyTuple *args, PyDict *kwargs)
 		previous_handler = std::get<PyObject *>(it->second);
 	}
 
-	if (std::signal(signalnum, sighandler) == SIG_ERR) {
+	if (signalnum >= NSIG) { return Err(value_error("signal number out of range")); }
+
+	static_assert(NSIG < std::numeric_limits<int>::max());
+	if (std::signal(static_cast<int>(signalnum), sighandler) == SIG_ERR) {
 		return Err(value_error("error setting signal handler"));
 	}
 
diff --git a/src/runtime/modules/sre/Pattern.cpp b/src/runtime/modules/sre/Pattern.cpp
@@ -8,11 +8,13 @@
 #include "runtime/PyObject.hpp"
 #include "runtime/PyTuple.hpp"
 #include "runtime/TypeError.hpp"
+#include "runtime/ValueError.hpp"
 #include "runtime/types/api.hpp"
 #include "runtime/types/builtin.hpp"
 #include "utilities.hpp"
 #include "vm/VM.hpp"
 #include <cstdint>
+#include <limits>
 
 using namespace py;
 using namespace py::sre;
@@ -49,7 +51,14 @@ PyResult<Pattern *> Pattern::create(PyObject *pattern,
 		auto el_ = PyObject::from(el);
 		if (el_.is_err()) { return Err(el_.unwrap_err()); }
 		if (!el_.unwrap()->type()->issubclass(types::integer())) { TODO(); }
-		code_vec.push_back(static_cast<const PyInteger &>(*el_.unwrap()).as_size_t());
+		auto value = static_cast<const PyInteger &>(*el_.unwrap()).as_size_t();
+		if (!fits_in<uint32_t>(value)) {
+			return Err(value_error("code value {} does not fit in [{}, {}]",
+				value,
+				std::numeric_limits<uint32_t>::min(),
+				std::numeric_limits<uint32_t>::max()));
+		}
+		code_vec.push_back(static_cast<uint32_t>(value));
 	}
 
 	std::optional<bool> isbytes;
diff --git a/src/runtime/modules/sre/module.cpp b/src/runtime/modules/sre/module.cpp
@@ -95,7 +95,7 @@ PyResult<PyObject *> ascii_iscased(PyTuple *args, PyDict *kwargs)
 
 	auto ch = character->as_big_int().get_si();
 	if (ch >= 128) { return Ok(py_false()); }
-	return Ok(std::isalpha(ch) ? py_true() : py_false());
+	return Ok(std::isalpha(static_cast<int>(ch)) ? py_true() : py_false());
 }
 
 PyResult<PyObject *> ascii_tolower(PyTuple *args, PyDict *kwargs)
@@ -111,7 +111,7 @@ PyResult<PyObject *> ascii_tolower(PyTuple *args, PyDict *kwargs)
 
 	auto ch = character->as_big_int().get_si();
 	if (ch >= 128) { return Ok(character); }
-	return PyInteger::create(std::tolower(ch));
+	return PyInteger::create(std::tolower(static_cast<int>(ch)));
 }
 
 
diff --git a/src/runtime/modules/time/module.cpp b/src/runtime/modules/time/module.cpp
@@ -56,10 +56,10 @@ PyModule *time_module()
 
 	module->add_symbol(PyString::create("monotonic").unwrap(),
 		PyNativeFunction::create("monotonic", [](PyTuple *, PyDict *) {
-			const double ns = std::chrono::duration_cast<std::chrono::nanoseconds>(
+			auto ns = std::chrono::duration_cast<std::chrono::nanoseconds>(
 				std::chrono::steady_clock::now().time_since_epoch())
-								  .count();
-			return PyFloat::create(ns / std::nano::den);
+						  .count();
+			return PyFloat::create(static_cast<double>(ns) / std::nano::den);
 		}).unwrap());
 
 	return module;
diff --git a/src/utilities.hpp b/src/utilities.hpp
@@ -83,3 +83,30 @@ template<class To, class From> constexpr To bit_cast(const From &from) noexcept
 	return std::bit_cast<To>(from);
 }
 #endif
+
+
+template<typename T>
+concept Integral = std::is_integral_v<T>;
+
+template<Integral Target, Integral Source> bool fits_in(Source value)
+{
+	constexpr bool source_is_signed = std::is_signed<Source>::value;
+	constexpr bool target_is_signed = std::is_signed<Target>::value;
+
+	// Case 1: Source is signed, Target is unsigned
+	if constexpr (source_is_signed && !target_is_signed) {
+		if (value < 0) { return false; }
+		// Now we know value is non-negative, safe to cast and compare
+		return static_cast<std::make_unsigned_t<Source>>(value)
+			   <= std::numeric_limits<Target>::max();
+	}
+	// Case 2: Source is unsigned, Target is signed
+	else if constexpr (!source_is_signed && target_is_signed) {
+		return value
+			   <= static_cast<std::make_unsigned_t<Target>>(std::numeric_limits<Target>::max());
+	} else {
+		// Case 3: Both signed or both unsigned
+		return value >= std::numeric_limits<Target>::min()
+			   && value <= std::numeric_limits<Target>::max();
+	}
+}

Original file line number	Diff line number	Diff line change
`@@ -466,21 +466,21 @@ class LinearScanRegisterAllocation`
`466`	`466`	`auto current_value = std::get<mlir::Value>(cur.value);`
`467`	`467`
`468`	`468`	`// Resolve block arguments to their defining operations`
`469`		`- if (current_value.isa<mlir::BlockArgument>()) {`
	`469`	`+ if (mlir::isa<mlir::BlockArgument>(current_value)) {`
`470`	`470`	`if (auto it = live_interval_analysis.block_input_mappings.find(cur.value);`
`471`	`471`	`it != live_interval_analysis.block_input_mappings.end()) {`
`472`	`472`	`for (auto mapped_value : it->second) {`
`473`	`473`	`ASSERT(!std::holds_alternative<ForwardedOutput>(mapped_value));`
`474`	`474`	`if (clobbers_r0(std::get<mlir::Value>(mapped_value))) {`
`475`		`- ASSERT(current_value.isa<mlir::BlockArgument>());`
	`475`	`+ ASSERT(mlir::isa<mlir::BlockArgument>(current_value));`
`476`	`476`	`current_value = std::get<mlir::Value>(mapped_value);`
`477`	`477`	`break;`
`478`	`478`	`}`
`479`	`479`	`}`
`480`	`480`	`}`
`481`	`481`	`}`
`482`	`482`
`483`		`- ASSERT(!current_value.isa<mlir::BlockArgument>());`
	`483`	`+ ASSERT(!mlir::isa<mlir::BlockArgument>(current_value));`
`484`	`484`	`auto loc = current_value.getLoc();`
`485`	`485`
`486`	`486`	`// Insert: push r{cur_reg}, move r{scratch}, r{cur_reg}, pop r{cur_reg}`
Original file line number	Diff line number	Diff line change
`@@ -505,7 +505,7 @@ PyResult<PyObject *> PyList::__mul__(size_t count) const`
`505`	`505`	`if (count <= 0) { return PyList::create(); }`
`506`	`506`	`std::vector<Value> values;`
`507`	`507`	`values.reserve(count * m_elements.size());`
`508`		`- for (auto _ : std::views::iota(size_t{ 0 }, count)) {`
	`508`	`+ for ([[maybe_unused]] auto _ : std::views::iota(size_t{ 0 }, count)) {`
`509`	`509`	`values.insert(values.end(), m_elements.begin(), m_elements.end());`
`510`	`510`	`}`
`511`	`511`
Original file line number	Diff line number	Diff line change
`@@ -95,7 +95,7 @@ PyResult<PyObject > ascii_iscased(PyTuple args, PyDict *kwargs)`
`95`	`95`
`96`	`96`	`auto ch = character->as_big_int().get_si();`
`97`	`97`	`if (ch >= 128) { return Ok(py_false()); }`
`98`		`- return Ok(std::isalpha(ch) ? py_true() : py_false());`
	`98`	`+ return Ok(std::isalpha(static_cast<int>(ch)) ? py_true() : py_false());`
`99`	`99`	`}`
`100`	`100`
`101`	`101`	`PyResult<PyObject > ascii_tolower(PyTuple args, PyDict *kwargs)`
`@@ -111,7 +111,7 @@ PyResult<PyObject > ascii_tolower(PyTuple args, PyDict *kwargs)`
`111`	`111`
`112`	`112`	`auto ch = character->as_big_int().get_si();`
`113`	`113`	`if (ch >= 128) { return Ok(character); }`
`114`		`- return PyInteger::create(std::tolower(ch));`
	`114`	`+ return PyInteger::create(std::tolower(static_cast<int>(ch)));`
`115`	`115`	`}`
`116`	`116`
`117`	`117`