[RFC] Case sensitive PHP

NEWS

@@ -3,6 +3,15 @@ PHP                                                                        NEWS
 ?? ??? ????, PHP 8.6.0alpha1
 
 - Core:
+  . PHP is now case-sensitive for function and class names. Calling a function
+    using incorrect casing (e.g. STRLEN() instead of strlen()) now raises a
+    fatal Error. (jorgsowa)
+  . Class, interface, trait, and enum names are now stored and looked up using
+    their canonical casing. Referencing them with incorrect casing now raises a
+    fatal Error across all language constructs and APIs that accept a class
+    name. (jorgsowa)
+  . Namespace segments in references and use imports are now matched
+    case-sensitively. Using incorrect casing raises a fatal Error. (jorgsowa)
   . Added first-class callable cache to share instances for the duration of the
     request. (ilutov)
   . It is now possible to use reference assign on WeakMap without the key

UPGRADING

@@ -19,6 +19,51 @@ PHP 8.6 UPGRADE NOTES
 1. Backward Incompatible Changes
 ========================================
 
+- Core:
+  . PHP is now case-sensitive for function and class names. The global function
+    table and class table are now keyed by the canonical (declared) name rather
+    than its lowercase equivalent. As a result, calling a function with
+    incorrect casing (e.g. STRLEN() instead of strlen()) now raises a fatal
+    Error ("Call to undefined function STRLEN()"), and referencing a class,
+    interface, trait, or enum with incorrect casing (e.g. new FOO() when the
+    class is declared as Foo) now raises a fatal Error ("Class 'FOO' not
+    found"). This applies to all language constructs and APIs that perform a
+    name lookup: new, instanceof, catch, extends, implements, use (trait), type
+    declarations, callable strings/arrays, class_exists() and related
+    introspection functions, Closure::bind()/bindTo(), Reflection constructors
+    and methods, and the SOAP classmap option.
+    RFC: https://wiki.php.net/rfc/case_sensitive_php
+  . Namespace segments in use imports are now matched case-sensitively. A use
+    import whose path does not match the canonical casing of the target
+    namespace segment will produce a fatal Error when the imported name is
+    first resolved.
+    RFC: https://wiki.php.net/rfc/case_sensitive_php
+  . Declaring a magic method with incorrect casing (e.g. __tostring() instead
+    of __toString()) is now a compile-time fatal Error. Previously magic
+    method names were matched case-insensitively; under case-sensitive method
+    names a wrong-cased declaration would otherwise silently lose its magic
+    behavior.
+    RFC: https://wiki.php.net/rfc/case_sensitive_php
+  . Attribute names are now matched case-sensitively, consistent with class
+    names. Internal attributes such as #[\Deprecated], #[\Override] or
+    #[\AllowDynamicProperties] are only recognized with their exact casing,
+    and the name filter of ReflectionFunctionAbstract::getAttributes() and
+    related methods compares case-sensitively. The "self", "parent" and
+    "static" keywords in callable strings remain case-insensitive.
+    RFC: https://wiki.php.net/rfc/case_sensitive_php
+  . The namespace portion of constant names is now case-sensitive, matching
+    the constant base name. Referencing a namespaced constant with incorrect
+    namespace casing (e.g. myapp\FOO when defined as MyApp\FOO) now raises
+    an "Undefined constant" Error, including via constant() and defined().
+    RFC: https://wiki.php.net/rfc/case_sensitive_php
+
+- SOAP:
+  . WSDL operation names are now matched case-sensitively when dispatching
+    requests and when calling operations through SoapClient. This matches
+    XML, where element names are case-sensitive. SoapServer::addFunction()
+    also requires the exact case of the registered PHP function.
+    RFC: https://wiki.php.net/rfc/case_sensitive_php
+
 - DOM:
   . Properties previously documented as @readonly (e.g. DOMNode::$nodeType,
     DOMDocument::$xmlEncoding, DOMEntity::$actualEncoding, ::$encoding,

Zend/Optimizer/compact_literals.c

@@ -134,43 +134,43 @@ void zend_optimizer_compact_literals(zend_op_array *op_array, zend_optimizer_ctx
 					LITERAL_INFO(opline->op1.constant, 1);
 					break;
 				case ZEND_INIT_FCALL_BY_NAME:
-					LITERAL_INFO(opline->op2.constant, 2);
+					LITERAL_INFO(opline->op2.constant, 1);
 					break;
 				case ZEND_INIT_NS_FCALL_BY_NAME:
-					LITERAL_INFO(opline->op2.constant, 3);
+					LITERAL_INFO(opline->op2.constant, 2);
 					break;
 				case ZEND_INIT_METHOD_CALL:
 					if (opline->op1_type == IS_CONST) {
 						LITERAL_INFO(opline->op1.constant, 1);
 					}
 					if (opline->op2_type == IS_CONST) {
-						LITERAL_INFO(opline->op2.constant, 2);
+						LITERAL_INFO(opline->op2.constant, 1);
 					}
 					break;
 				case ZEND_INIT_STATIC_METHOD_CALL:
 					if (opline->op1_type == IS_CONST) {
-						LITERAL_INFO(opline->op1.constant, 2);
+						LITERAL_INFO(opline->op1.constant, 1);
 					}
 					if (opline->op2_type == IS_CONST) {
-						LITERAL_INFO(opline->op2.constant, 2);
+						LITERAL_INFO(opline->op2.constant, 1);
 					}
 					break;
 				case ZEND_INIT_PARENT_PROPERTY_HOOK_CALL:
 					LITERAL_INFO(opline->op1.constant, 1);
 					break;
 				case ZEND_CATCH:
-					LITERAL_INFO(opline->op1.constant, 2);
+					LITERAL_INFO(opline->op1.constant, 1);
 					break;
 				case ZEND_FETCH_CONSTANT:
 					if (opline->op1.num & IS_CONSTANT_UNQUALIFIED_IN_NAMESPACE) {
-						LITERAL_INFO(opline->op2.constant, 3);
-					} else {
 						LITERAL_INFO(opline->op2.constant, 2);
+					} else {
+						LITERAL_INFO(opline->op2.constant, 1);
 					}
 					break;
 				case ZEND_FETCH_CLASS_CONSTANT:
 					if (opline->op1_type == IS_CONST) {
-						LITERAL_INFO(opline->op1.constant, 2);
+						LITERAL_INFO(opline->op1.constant, 1);
 					}
 					if (opline->op2_type == IS_CONST) {
 						LITERAL_INFO(opline->op2.constant, 1);
@@ -192,7 +192,7 @@ void zend_optimizer_compact_literals(zend_op_array *op_array, zend_optimizer_ctx
 				case ZEND_POST_DEC_STATIC_PROP:
 				case ZEND_ASSIGN_STATIC_PROP_OP:
 					if (opline->op2_type == IS_CONST) {
-						LITERAL_INFO(opline->op2.constant, 2);
+						LITERAL_INFO(opline->op2.constant, 1);
 					}
 					if (opline->op1_type == IS_CONST) {
 						LITERAL_INFO(opline->op1.constant, 1);
@@ -201,12 +201,12 @@ void zend_optimizer_compact_literals(zend_op_array *op_array, zend_optimizer_ctx
 				case ZEND_FETCH_CLASS:
 				case ZEND_INSTANCEOF:
 					if (opline->op2_type == IS_CONST) {
-						LITERAL_INFO(opline->op2.constant, 2);
+						LITERAL_INFO(opline->op2.constant, 1);
 					}
 					break;
 				case ZEND_NEW:
 					if (opline->op1_type == IS_CONST) {
-						LITERAL_INFO(opline->op1.constant, 2);
+						LITERAL_INFO(opline->op1.constant, 1);
 					}
 					break;
 				case ZEND_DECLARE_CLASS:

Zend/Optimizer/dfa_pass.c

@@ -277,9 +277,7 @@ static inline bool can_elide_list_type(
 			return can_elide_list_type(script, op_array, use_info, *single_type);
 		}
 		if (ZEND_TYPE_HAS_NAME(*single_type)) {
-			zend_string *lcname = zend_string_tolower(ZEND_TYPE_NAME(*single_type));
-			const zend_class_entry *ce = zend_optimizer_get_class_entry(script, op_array, lcname);
-			zend_string_release(lcname);
+			const zend_class_entry *ce = zend_optimizer_get_class_entry(script, op_array, ZEND_TYPE_NAME(*single_type));
 			bool result = ce && safe_instanceof(use_info->ce, ce);
 			if (result == !is_intersection) {
 				return result;
@@ -410,7 +408,7 @@ static uint32_t zend_dfa_optimize_calls(zend_op_array *op_array, zend_ssa *ssa)
 			if ((op->opcode == ZEND_FRAMELESS_ICALL_2
 			  || (op->opcode == ZEND_FRAMELESS_ICALL_3 && (op + 1)->op1_type == IS_CONST))
 			 && call_info->callee_func
-			 && zend_string_equals_literal_ci(call_info->callee_func->common.function_name, "in_array")) {
+			 && zend_string_equals_literal(call_info->callee_func->common.function_name, "in_array")) {
 				bool strict = false;
 				bool has_opdata = op->opcode == ZEND_FRAMELESS_ICALL_3;
 				ZEND_ASSERT(!call_info->is_prototype);

Zend/Optimizer/optimize_func_calls.c

@@ -200,14 +200,19 @@ void zend_optimize_func_calls(zend_op_array *op_array, zend_optimizer_ctx *ctx)
 					} else if (fcall->opcode == ZEND_INIT_FCALL_BY_NAME) {
 						fcall->opcode = ZEND_INIT_FCALL;
 						fcall->op1.num = zend_vm_calc_used_stack(fcall->extended_value, call_stack[call].func);
-						literal_dtor(&ZEND_OP2_LITERAL(fcall));
-						fcall->op2.constant = fcall->op2.constant + 1;
 					} else if (fcall->opcode == ZEND_INIT_NS_FCALL_BY_NAME) {
 						fcall->opcode = ZEND_INIT_FCALL;
 						fcall->op1.num = zend_vm_calc_used_stack(fcall->extended_value, call_stack[call].func);
-						literal_dtor(&op_array->literals[fcall->op2.constant]);
-						literal_dtor(&op_array->literals[fcall->op2.constant + 2]);
-						fcall->op2.constant = fcall->op2.constant + 1;
+						/* slot 0 = ns-qualified name, slot 1 = unqualified fallback */
+						zval *ns_slot = &op_array->literals[fcall->op2.constant];
+						zval *unq_slot = &op_array->literals[fcall->op2.constant + 1];
+						if (!zend_hash_find(EG(function_table), Z_STR_P(ns_slot))
+								&& Z_TYPE_P(unq_slot) == IS_STRING) {
+							literal_dtor(ns_slot);
+							fcall->op2.constant++;
+						} else {
+							literal_dtor(&op_array->literals[fcall->op2.constant + 1]);
+						}
 					} else if (fcall->opcode == ZEND_INIT_STATIC_METHOD_CALL
 							|| fcall->opcode == ZEND_INIT_METHOD_CALL
 							|| fcall->opcode == ZEND_INIT_PARENT_PROPERTY_HOOK_CALL

Zend/Optimizer/pass1.c

@@ -220,7 +220,7 @@ void zend_optimizer_pass1(zend_op_array *op_array, zend_optimizer_ctx *ctx)
 			}
 
 			/* define("name", scalar); */
-			if (zend_string_equals_literal_ci(Z_STR(ZEND_OP2_LITERAL(init_opline)), "define")) {
+			if (zend_string_equals_literal(Z_STR(ZEND_OP2_LITERAL(init_opline)), "define")) {
 
 				if (Z_TYPE(ZEND_OP1_LITERAL(send1_opline)) == IS_STRING && send2_opline) {
 

Zend/Optimizer/zend_inference.c

@@ -2387,10 +2387,8 @@ static uint32_t zend_convert_type(const zend_script *script, zend_type type, zen
 			/* As we only have space to store one CE,
 			 * we use a plain object type for class unions. */
 			if (ZEND_TYPE_HAS_NAME(type)) {
-				zend_string *lcname = zend_string_tolower(ZEND_TYPE_NAME(type));
 				// TODO: Pass through op_array.
-				*pce = zend_optimizer_get_class_entry(script, NULL, lcname);
-				zend_string_release_ex(lcname, 0);
+				*pce = zend_optimizer_get_class_entry(script, NULL, ZEND_TYPE_NAME(type));
 			}
 		}
 	}
@@ -2477,7 +2475,7 @@ static const zend_property_info *zend_fetch_static_prop_info(const zend_script *
 			}
 		} else if (opline->op2_type == IS_CONST) {
 			const zval *zv = CRT_CONSTANT(opline->op2);
-			ce = zend_optimizer_get_class_entry(script, op_array, Z_STR_P(zv + 1));
+			ce = zend_optimizer_get_class_entry(script, op_array, Z_STR_P(zv));
 		}
 
 		if (ce) {
@@ -3369,7 +3367,7 @@ static zend_always_inline zend_result _zend_update_type_info(
 			} else if (opline->op2_type == IS_CONST) {
 				zval *zv = CRT_CONSTANT(opline->op2);
 				if (Z_TYPE_P(zv) == IS_STRING) {
-					ce = zend_optimizer_get_class_entry(script, op_array, Z_STR_P(zv+1));
+					ce = zend_optimizer_get_class_entry(script, op_array, Z_STR_P(zv));
 					UPDATE_SSA_OBJ_TYPE(ce, 0, ssa_op->result_def);
 				} else {
 					UPDATE_SSA_OBJ_TYPE(NULL, 0, ssa_op->result_def);