ext/intl: Refactor global error resets (#22990)

Introduce PHP_INTL_FUNCTION_WITH_ERROR_RESET macro to standardize
and automate the global Intl error state reset at the start of
every userland PHP_FUNCTION. Also change the IC_METHOD function
to automatically reset error state in the start of the defined function.

The macro wraps the function entry point, performs
intl_error_reset(NULL), and delegates the actual logic to a
separate _impl function. This reduces boilerplate and prevents
bugs caused by missing error resets, as seen in GH-22931 and
GH-22500.

This refactor applies to all functions with a manual reset call
and does not affect internal helper functions that handle errors
independently.

Author: LamentXU123

ext/intl/ERROR_CONVENTIONS.md

@@ -93,13 +93,21 @@ void intl_error_reset(NULL);             /* reset global error */
 void intl_errors_reset(intl_error* err); /* reset global and object error */
 ```
 
-In practice, `intl_errors_reset()` is not used because most classes have also
-plain functions mapped to the same internal functions as their instance methods.
-Fetching of the object is done with `zend_parse_method_parameters()` instead of
-directly using `getThis()`. Therefore, no reference to object is obtained until
-the arguments are fully parsed. Without a reference to the object, there's no
-way to reset the object's internal error code. Instead, resetting of the
-object's internal error code is done upon fetching the object from its zval.
+Procedural functions that reset the global error should normally use
+`PHP_INTL_FUNCTION_WITH_ERROR_RESET(name)`, which resets the global error before
+entering the implementation body.
+Class-specific method helper macros may use the same pattern when all methods
+covered by the macro reset the global error.
+
+`intl_errors_reset()` may be used directly when the object is available before
+argument parsing, for example in methods that only operate on `ZEND_THIS()`.
+For classes that have plain functions mapped to the same internal functions as
+their instance methods, fetching of the object is done with
+`zend_parse_method_parameters()` instead of directly using `getThis()`.
+Therefore, no reference to object is obtained until the arguments are fully
+parsed. Without a reference to the object, there's no way to reset the object's
+internal error code. Instead, resetting of the object's internal error code is
+done upon fetching the object from its zval.
 
 Example:
 

ext/intl/calendar/calendar_methods.cpp

@@ -70,14 +70,13 @@ U_CFUNC PHP_METHOD(IntlCalendar, __construct)
 		0 );
 }
 
-U_CFUNC PHP_FUNCTION(intlcal_create_instance)
+PHP_INTL_FUNCTION_WITH_ERROR_RESET(intlcal_create_instance)
 {
 	zend_object *timezone_object = nullptr;
 	zend_string *timezone_string = nullptr;
 	char	        *locale_str	= NULL;
 	size_t		locale_len      = 0;
 	UErrorCode	status			= U_ZERO_ERROR;
-	intl_error_reset(NULL);
 
 	ZEND_PARSE_PARAMETERS_START(0, 2)
 		Z_PARAM_OPTIONAL
@@ -158,15 +157,14 @@ class BugStringCharEnumeration : public StringEnumeration
 };
 UOBJECT_DEFINE_RTTI_IMPLEMENTATION(BugStringCharEnumeration)
 
-U_CFUNC PHP_FUNCTION(intlcal_get_keyword_values_for_locale)
+PHP_INTL_FUNCTION_WITH_ERROR_RESET(intlcal_get_keyword_values_for_locale)
 {
 	UErrorCode	status = U_ZERO_ERROR;
 	char		*key,
 				*locale;
 	size_t			key_len,
 				locale_len;
 	bool	commonly_used;
-	intl_error_reset(NULL);
 
 	ZEND_PARSE_PARAMETERS_START(3, 3)
 		Z_PARAM_STRING(key, key_len)
@@ -186,19 +184,15 @@ U_CFUNC PHP_FUNCTION(intlcal_get_keyword_values_for_locale)
 	IntlIterator_from_StringEnumeration(se, return_value);
 }
 
-U_CFUNC PHP_FUNCTION(intlcal_get_now)
+PHP_INTL_FUNCTION_WITH_ERROR_RESET(intlcal_get_now)
 {
-	intl_error_reset(NULL);
-
 	ZEND_PARSE_PARAMETERS_NONE();
 
 	RETURN_DOUBLE((double)Calendar::getNow());
 }
 
-U_CFUNC PHP_FUNCTION(intlcal_get_available_locales)
+PHP_INTL_FUNCTION_WITH_ERROR_RESET(intlcal_get_available_locales)
 {
-	intl_error_reset(NULL);
-
 	ZEND_PARSE_PARAMETERS_NONE();
 
 	int32_t count;
@@ -1018,7 +1012,7 @@ U_CFUNC PHP_FUNCTION(intlcal_set_skipped_wall_time_option)
 	RETURN_TRUE;
 }
 
-U_CFUNC PHP_FUNCTION(intlcal_from_date_time)
+PHP_INTL_FUNCTION_WITH_ERROR_RESET(intlcal_from_date_time)
 {
 	zend_object     *date_obj;
 	zend_string     *date_str;
@@ -1029,7 +1023,6 @@ U_CFUNC PHP_FUNCTION(intlcal_from_date_time)
 	TimeZone		*timeZone;
 	UErrorCode		status = U_ZERO_ERROR;
 	Calendar        *cal;
-	intl_error_reset(NULL);
 
 	ZEND_PARSE_PARAMETERS_START(1, 2)
 		Z_PARAM_OBJ_OF_CLASS_OR_STR(date_obj, php_date_get_date_ce(), date_str)

ext/intl/locale/locale_methods.cpp

@@ -836,7 +836,7 @@ U_CFUNC PHP_FUNCTION(locale_get_display_keyword_value)
  /* {{{ return an associative array containing keyword-value
  * pairs for this locale. The keys are keys to the array (doh!)
  */
-U_CFUNC PHP_FUNCTION( locale_get_keywords )
+PHP_INTL_FUNCTION_WITH_ERROR_RESET(locale_get_keywords)
 {
 	UEnumeration*   e        = NULL;
 	UErrorCode      status   = U_ZERO_ERROR;
@@ -847,7 +847,6 @@ U_CFUNC PHP_FUNCTION( locale_get_keywords )
 	char*       	        loc_name        = NULL;
 	size_t        	 	loc_name_len    = 0;
 
-	intl_error_reset( NULL );
 
 	ZEND_PARSE_PARAMETERS_START(1, 1)
 		Z_PARAM_PATH(loc_name, loc_name_len)
@@ -1051,15 +1050,14 @@ static int handleAppendResult( int result, smart_str* loc_name)
 * }}} */
 /* {{{ Creates a locale by combining the parts of locale-ID passed
 * }}} */
-U_CFUNC PHP_FUNCTION(locale_compose)
+PHP_INTL_FUNCTION_WITH_ERROR_RESET(locale_compose)
 {
 	smart_str      	loc_name_s = {NULL, 0};
 	smart_str *loc_name = &loc_name_s;
 	zval*			arr	= NULL;
 	HashTable*		hash_arr = NULL;
 	int 			result = 0;
 
-	intl_error_reset( NULL );
 
 	ZEND_PARSE_PARAMETERS_START(1, 1)
 		Z_PARAM_ARRAY(arr)
@@ -1232,13 +1230,12 @@ static int add_array_entry(const char* loc_name, zval* hash_arr, const char* key
 /* }}} */
 
 /* {{{ parses a locale-id into an array the different parts of it */
-U_CFUNC PHP_FUNCTION(locale_parse)
+PHP_INTL_FUNCTION_WITH_ERROR_RESET(locale_parse)
 {
 	char*          loc_name        = NULL;
 	size_t         loc_name_len    = 0;
 	int         grOffset    	= 0;
 
-	intl_error_reset( NULL );
 
 	ZEND_PARSE_PARAMETERS_START(1, 1)
 		Z_PARAM_PATH(loc_name, loc_name_len)
@@ -1268,7 +1265,7 @@ U_CFUNC PHP_FUNCTION(locale_parse)
 /* }}} */
 
 /* {{{ gets an array containing the list of variants, or null */
-U_CFUNC PHP_FUNCTION(locale_get_all_variants)
+PHP_INTL_FUNCTION_WITH_ERROR_RESET(locale_get_all_variants)
 {
 	char*  	                loc_name        = NULL;
 	size_t    		loc_name_len    = 0;
@@ -1278,7 +1275,6 @@ U_CFUNC PHP_FUNCTION(locale_get_all_variants)
 	zend_string*	variant		= NULL;
 	char*	saved_ptr	= NULL;
 
-	intl_error_reset( NULL );
 
 	ZEND_PARSE_PARAMETERS_START(1, 1)
 		Z_PARAM_PATH(loc_name, loc_name_len)
@@ -1352,7 +1348,7 @@ static int strToMatch(const char* str ,char *retstr)
 /* {{{ Checks if a $langtag filter matches with $locale according to RFC 4647's basic filtering algorithm */
 /* }}} */
 /* {{{ Checks if a $langtag filter matches with $locale according to RFC 4647's basic filtering algorithm */
-U_CFUNC PHP_FUNCTION(locale_filter_matches)
+PHP_INTL_FUNCTION_WITH_ERROR_RESET(locale_filter_matches)
 {
 	char*       	lang_tag        = NULL;
 	size_t         	lang_tag_len    = 0;
@@ -1372,7 +1368,6 @@ U_CFUNC PHP_FUNCTION(locale_filter_matches)
 	bool 	boolCanonical 	= 0;
 	UErrorCode	status		= U_ZERO_ERROR;
 
-	intl_error_reset( NULL );
 
 	ZEND_PARSE_PARAMETERS_START(2, 3)
 		Z_PARAM_PATH(lang_tag, lang_tag_len)
@@ -1640,7 +1635,7 @@ static zend_string* lookup_loc_range(const char* loc_range, HashTable* hash_arr,
 /* {{{ Searches the items in $langtag for the best match to the language
 * range
 */
-U_CFUNC PHP_FUNCTION(locale_lookup)
+PHP_INTL_FUNCTION_WITH_ERROR_RESET(locale_lookup)
 {
 	zend_string*   	fallback_loc_str	= NULL;
 	char*    	loc_range      		= NULL;
@@ -1651,7 +1646,6 @@ U_CFUNC PHP_FUNCTION(locale_lookup)
 	bool	boolCanonical	= 0;
 	zend_string* 	result_str	= NULL;
 
-	intl_error_reset( NULL );
 
 	ZEND_PARSE_PARAMETERS_START(2, 4)
 		Z_PARAM_ARRAY(arr)

ext/intl/normalizer/normalizer_normalize.cpp

@@ -73,7 +73,7 @@ static UBool intl_is_normalized(zend_long form, const UChar *uinput, int32_t uin
 }/*}}}*/
 
 /* {{{ Normalize a string. */
-U_CFUNC PHP_FUNCTION( normalizer_normalize )
+PHP_INTL_FUNCTION_WITH_ERROR_RESET(normalizer_normalize)
 {
 	char*			input = NULL;
 	/* form is optional, defaults to FORM_C */
@@ -92,7 +92,6 @@ U_CFUNC PHP_FUNCTION( normalizer_normalize )
 
 	int32_t			size_needed;
 
-	intl_error_reset( NULL );
 
 	/* Parse parameters. */
 	if( zend_parse_method_parameters( ZEND_NUM_ARGS(), getThis(), "s|l",
@@ -202,7 +201,7 @@ U_CFUNC PHP_FUNCTION( normalizer_normalize )
 /* }}} */
 
 /* {{{ Test if a string is in a given normalization form. */
-U_CFUNC PHP_FUNCTION( normalizer_is_normalized )
+PHP_INTL_FUNCTION_WITH_ERROR_RESET(normalizer_is_normalized)
 {
 	char*	 	input = NULL;
 	/* form is optional, defaults to FORM_C */
@@ -215,7 +214,6 @@ U_CFUNC PHP_FUNCTION( normalizer_is_normalized )
 
 	UBool		uret = false;
 
-	intl_error_reset( NULL );
 
 	/* Parse parameters. */
 	if( zend_parse_method_parameters( ZEND_NUM_ARGS(), getThis(), "s|l",
@@ -278,7 +276,7 @@ U_CFUNC PHP_FUNCTION( normalizer_is_normalized )
 /* }}} */
 
 /* {{{ Returns the Decomposition_Mapping property for the given UTF-8 encoded code point. */
-U_CFUNC PHP_FUNCTION( normalizer_get_raw_decomposition )
+PHP_INTL_FUNCTION_WITH_ERROR_RESET(normalizer_get_raw_decomposition)
 {
 	char* input = NULL;
 	size_t input_length = 0;
@@ -293,7 +291,6 @@ U_CFUNC PHP_FUNCTION( normalizer_get_raw_decomposition )
 
 	zend_long form = NORMALIZER_DEFAULT;
 
-	intl_error_reset(NULL);
 
 	ZEND_PARSE_PARAMETERS_START(1, 2)
 		Z_PARAM_STRING(input, input_length)

ext/intl/php_intl.h

@@ -71,6 +71,15 @@ PHP_MINFO_FUNCTION(intl);
 const char *intl_locale_get_default( void );
 char *canonicalize_locale_string(const char* locale);
 
+#define PHP_INTL_FUNCTION_WITH_ERROR_RESET(name) \
+	static void php_intl_##name##_impl(INTERNAL_FUNCTION_PARAMETERS); \
+	U_CFUNC PHP_FUNCTION(name) \
+	{ \
+		intl_error_reset(NULL); \
+		php_intl_##name##_impl(INTERNAL_FUNCTION_PARAM_PASSTHRU); \
+	} \
+	static void php_intl_##name##_impl(INTERNAL_FUNCTION_PARAMETERS)
+
 #define PHP_INTL_VERSION PHP_VERSION
 
 #endif  /* PHP_INTL_H */

ext/intl/resourcebundle/resourcebundle_class.cpp

@@ -333,7 +333,7 @@ U_CFUNC PHP_FUNCTION( resourcebundle_count )
 }
 
 /* {{{ Get available locales from ResourceBundle name */
-U_CFUNC PHP_FUNCTION( resourcebundle_locales )
+PHP_INTL_FUNCTION_WITH_ERROR_RESET(resourcebundle_locales)
 {
 	char * bundlename;
 	size_t    bundlename_len = 0;
@@ -343,8 +343,6 @@ U_CFUNC PHP_FUNCTION( resourcebundle_locales )
 	UEnumeration *icuenum;
 	UErrorCode   icuerror = U_ZERO_ERROR;
 
-	intl_errors_reset( NULL );
-
 	ZEND_PARSE_PARAMETERS_START(1, 1)
 		Z_PARAM_STRING(bundlename, bundlename_len)
 	ZEND_PARSE_PARAMETERS_END();