Merge branch 'PHP-8.5'

* PHP-8.5:
  uri: Update to uriparser-0.9.9-21-g08df3b2 (#19992)

Author: TimWolla

ext/uri/config.m4

@@ -33,7 +33,7 @@ if test "$PHP_EXTERNAL_URIPARSER" = "no"; then
   $URIPARSER_DIR/src/UriSetScheme.c $URIPARSER_DIR/src/UriSetUserInfo.c $URIPARSER_DIR/src/UriShorten.c $URIPARSER_DIR/src/UriVersion.c"
   URI_CFLAGS="-DURI_STATIC_BUILD"
 else
-  PKG_CHECK_MODULES([LIBURIPARSER], [liburiparser >= 0.9.9])
+  PKG_CHECK_MODULES([LIBURIPARSER], [liburiparser >= 0.9.10])
   PHP_EVAL_LIBLINE([$LIBURIPARSER_LIBS], [URI_SHARED_LIBADD])
   PHP_EVAL_INCLINE([$LIBURIPARSER_CFLAGS])
 fi

ext/uri/uriparser/include/uriparser/Uri.h

@@ -137,7 +137,7 @@ typedef struct URI_TYPE(HostDataStruct) {
 	URI_TYPE(TextRange) ipFuture; /**< IPvFuture address
 		@note
 		With non-<c>NULL</c> members in UriUriStructA.hostData context,
-		this text range's pointers must be <em>idential</em> to those
+		this text range's pointers must be <em>identical</em> to those
 		of UriUriStructA.hostText at all times. */
 } URI_TYPE(HostData); /**< @copydoc UriHostDataStructA */
 
@@ -1420,7 +1420,7 @@ URI_PUBLIC UriBool URI_FUNC(IsWellFormedHostRegName)(const URI_CHAR * first, con
  *
  * @param first      <b>IN</b>: Pointer to first character
  * @param afterLast  <b>IN</b>: Pointer to character after the last one still in
- * @param hasHost    <b>IN</b>: Wether the target %URI has a non-<c>NULL</c> host set or not
+ * @param hasHost    <b>IN</b>: Whether the target %URI has a non-<c>NULL</c> host set or not
  * @return           <c>URI_TRUE</c> if non-<c>NULL</c> and well-formed, else <c>URI_FALSE</c>
  *
  * @see uriIsWellFormedFragmentA
@@ -2415,7 +2415,7 @@ URI_PUBLIC int URI_FUNC(SetUserInfoMm)(URI_TYPE(Uri) * uri,
  *       or even fully custom patches. As a result, the version string
  *       returned serves as nothing more than "based on that version",
  *       it does not guarantee equivalence to vanilla upstream releases
- *       or absence of additinal downstream patches.
+ *       or absence of additional downstream patches.
  *       It is nothing more than "a hint" and MUST NEVER be used to
  *       make decisions on in application code at runtime.
  *

ext/uri/uriparser/include/uriparser/UriBase.h

@@ -187,7 +187,7 @@ typedef struct UriIp6Struct {
 } UriIp6; /**< @copydoc UriIp6Struct */
 
 
-struct UriMemoryManagerStruct;  /* foward declaration to break loop */
+struct UriMemoryManagerStruct;  /* forward declaration to break loop */
 
 
 /**
@@ -287,28 +287,34 @@ typedef enum UriResolutionOptionsEnum {
 
 
 /**
- * Wraps a memory manager backend that only provides malloc and free
- * to make a complete memory manager ready to be used.
+ * Wraps a memory manager backend that only provides <c>malloc(3)</c> and
+ * <c>free(3)</c> to make a complete memory manager ready to be used.
  *
  * The core feature of this wrapper is that you don't need to implement
- * realloc if you don't want to.  The wrapped memory manager uses
- * backend->malloc, memcpy, and backend->free and soieof(size_t) extra
- * bytes per allocation to emulate fallback realloc for you.
+ * <c>realloc(3)</c> if you don't want to.  The wrapped memory manager uses
+ * <c>backend-&gt;malloc</c>, <c>memcpy(3)</c>, and <c>backend-&gt;free</c> and
+ * (at least) <c>sizeof(size_t)</c> extra bytes per allocation to emulate
+ * fallback <c>realloc(3)</c> for you.
  *
- * memory->calloc is uriEmulateCalloc.
- * memory->free uses backend->free and handles the size header.
- * memory->malloc uses backend->malloc and adds a size header.
- * memory->realloc uses memory->malloc, memcpy, and memory->free and reads
- *                 the size header.
- * memory->reallocarray is uriEmulateReallocarray.
+ * <ul>
+ * <li><c>memory-&gt;calloc</c> is <c>uriEmulateCalloc</c>.</li>
+ * <li><c>memory-&gt;free</c> uses <c>backend-&gt;free</c>,
+ *     and handles the size header.</li>
+ * <li><c>memory-&gt;malloc</c> uses <c>backend-&gt;malloc</c>,
+ *     and adds a size header.</li>
+ * <li><c>memory-&gt;realloc</c> uses <c>memory-&gt;malloc</c>,
+ *     <c>memcpy(3)</c> and <c>memory-&gt;free</c>,
+ *     and reads the size header.</li>
+ * <li><c>memory-&gt;reallocarray</c> is <c>uriEmulateReallocarray</c>.</li>
+ * </ul>
  *
- * The internal workings behind memory->free, memory->malloc, and
- * memory->realloc may change so the functions exposed by these function
- * pointer sshould be consided internal and not public API.
+ * The internal workings behind <c>memory-&gt;free</c>, <c>memory-&gt;malloc</c>,
+ * and <c>memory-&gt;realloc</c> may change, and the functions exposed by these
+ * function pointers should be considered internal and not public API.
  *
  * @param memory   <b>OUT</b>: Where to write the wrapped memory manager to
  * @param backend  <b>IN</b>: Memory manager to use as a backend
- * @return          Error code or 0 on success
+ * @return         Error code or 0 on success
  *
  * @see uriEmulateCalloc
  * @see uriEmulateReallocarray
@@ -321,7 +327,7 @@ URI_PUBLIC int uriCompleteMemoryManager(UriMemoryManager * memory,
 
 
 /**
- * Offers emulation of calloc(3) based on memory->malloc and memset.
+ * Offers emulation of calloc(3) based on memory-&gt;malloc and memset.
  * See "man 3 calloc" as well.
  *
  * @param memory  <b>IN</b>: Memory manager to use, should not be NULL
@@ -340,11 +346,11 @@ URI_PUBLIC void * uriEmulateCalloc(UriMemoryManager * memory,
 
 
 /**
- * Offers emulation of reallocarray(3) based on memory->realloc.
+ * Offers emulation of reallocarray(3) based on memory-&gt;realloc.
  * See "man 3 reallocarray" as well.
  *
  * @param memory  <b>IN</b>: Memory manager to use, should not be NULL
- * @param ptr     <b>IN</b>: Pointer allocated using memory->malloc/... or NULL
+ * @param ptr     <b>IN</b>: Pointer allocated using memory-&gt;malloc/... or NULL
  * @param nmemb   <b>IN</b>: Number of elements to allocate
  * @param size    <b>IN</b>: Size in bytes per element
  * @return        Pointer to allocated memory or NULL
@@ -369,18 +375,52 @@ URI_PUBLIC void * uriEmulateReallocarray(UriMemoryManager * memory,
  * 5. and frees that memory.
  *
  * It is recommended to compile with AddressSanitizer enabled
- * to take full advantage of uriTestMemoryManager.
+ * to take full advantage of <c>uriTestMemoryManager</c>.
+ *
+ * For backwards-compatibility, <c>uriTestMemoryManager</c>
+ * does not challenge pointer alignment; please see
+ * <c>uriTestMemoryManagerEx</c> for that feature.
  *
  * @param memory  <b>IN</b>: Memory manager to use, should not be NULL
  * @return        Error code or 0 on success
  *
  * @see uriEmulateCalloc
  * @see uriEmulateReallocarray
  * @see UriMemoryManager
+ * @see uriTestMemoryManagerEx
  * @since 0.9.0
  */
 URI_PUBLIC int uriTestMemoryManager(UriMemoryManager * memory);
 
 
 
+/**
+ * Run multiple tests against a given memory manager.
+ * For example, one test
+ * 1. allocates a small amount of memory,
+ * 2. writes some magic bytes to it,
+ * 3. reallocates it,
+ * 4. checks that previous values are still present,
+ * 5. and frees that memory.
+ *
+ * It is recommended to compile with both AddressSanitizer and
+ * UndefinedBehaviorSanitizer enabled to take full advantage of
+ * <c>uriTestMemoryManagerEx</c>. Note that environment variable
+ * <c>UBSAN_OPTIONS</c> may need adjustment to make UndefinedBehaviorSanitizer
+ * fatal (which by default it is not).
+ *
+ * @param memory              <b>IN</b>: Memory manager to use, should not be NULL
+ * @param challengeAlignment  <b>IN</b>: Whether to challenge pointer alignment
+ * @return                    Error code or 0 on success
+ *
+ * @see uriEmulateCalloc
+ * @see uriEmulateReallocarray
+ * @see UriMemoryManager
+ * @see uriTestMemoryManager
+ * @since 0.9.10
+ */
+URI_PUBLIC int uriTestMemoryManagerEx(UriMemoryManager * memory, UriBool challengeAlignment);
+
+
+
 #endif /* URI_BASE_H */

ext/uri/uriparser/src/UriCommon.c

@@ -208,14 +208,25 @@ UriBool URI_FUNC(RemoveDotSegmentsEx)(URI_TYPE(Uri) * uri,
 				 *
 				 * For example, changing "./http://foo" into "http://foo" would change semantics
 				 * and hence the dot segment is essential to that case and cannot be removed.
+				 *
+				 * Other examples that would change semantics are:
+				 * - cutting "/.//" down to "//"
+				 * - cutting "scheme:/.//" down to "scheme://".
 				 */
 				removeSegment = URI_TRUE;
-				if (relative && (walker == uri->pathHead) && (walker->next != NULL)) {
-					const URI_CHAR * ch = walker->next->text.first;
-					for (; ch < walker->next->text.afterLast; ch++) {
-						if (*ch == _UT(':')) {
-							removeSegment = URI_FALSE;
-							break;
+				if ((walker == uri->pathHead) && (walker->next != NULL)) {
+					/* Detect case "/.//" (with or without scheme) */
+					if ((walker->next->text.first == walker->next->text.afterLast)
+							&& (URI_FUNC(HasHost)(uri) == URI_FALSE)) {
+						removeSegment = URI_FALSE;
+					/* Detect case "./withcolon:" */
+					} else if (relative) {
+						const URI_CHAR * ch = walker->next->text.first;
+						for (; ch < walker->next->text.afterLast; ch++) {
+							if (*ch == _UT(':')) {
+								removeSegment = URI_FALSE;
+								break;
+							}
 						}
 					}
 				}
@@ -358,7 +369,7 @@ UriBool URI_FUNC(RemoveDotSegmentsEx)(URI_TYPE(Uri) * uri,
 								}
 								memory->free(memory, walker);
 							} else {
-								/* Re-use segment for "" path segment to represent trailing slash, update tail */
+								/* Reuse segment for "" path segment to represent trailing slash, update tail */
 								URI_TYPE(PathSegment) * const segment = walker;
 								if (pathOwned && (segment->text.first != segment->text.afterLast)) {
 									memory->free(memory, (URI_CHAR *)segment->text.first);
@@ -403,7 +414,7 @@ UriBool URI_FUNC(RemoveDotSegmentsEx)(URI_TYPE(Uri) * uri,
 								 * NEW: tail -> NULL */
 								uri->pathTail = NULL;
 							} else {
-								/* Re-use segment for "" path segment to represent trailing slash,
+								/* Reuse segment for "" path segment to represent trailing slash,
 								 * then update head and tail */
 								if (pathOwned && (walker->text.first != walker->text.afterLast)) {
 									memory->free(memory, (URI_CHAR *)walker->text.first);
@@ -696,7 +707,7 @@ static UriBool URI_FUNC(PrependNewDotSegment)(URI_TYPE(Uri) * uri, UriMemoryMana
 /* When dropping a scheme from a URI without a host and with a colon (":")
  * in the first path segment, a consecutive reparse would rightfully
  * mis-classify the first path segment as a scheme due to the colon.
- * To protect against this case, we prepend an artifical "." segment
+ * To protect against this case, we prepend an artificial "." segment
  * to the path in here; the function is called after the scheme has
  * just been dropped.
  *
@@ -751,7 +762,7 @@ UriBool URI_FUNC(FixPathNoScheme)(URI_TYPE(Uri) * uri,
 /* When dropping a host from a URI without a scheme, an absolute path
  * and and empty first path segment, a consecutive reparse would rightfully
  * mis-classify the first path segment as a host marker due to the "//".
- * To protect against this case, we prepend an artifical "." segment
+ * To protect against this case, we prepend an artificial "." segment
  * to the path in here; the function is called after the host has
  * just been dropped.
  *

ext/uri/uriparser/src/UriMemory.c

@@ -64,6 +64,15 @@
 
 
 
+#define URI_MAX(a, b)         (((a) > (b)) ? (a) : (b))
+
+/* NOTE: This intends to mimic MALLOC_ALIGNMENT of glibc */
+#define URI_MALLOC_ALIGNMENT  URI_MAX(2 * sizeof(size_t), sizeof(long double))
+
+#define URI_MALLOC_PADDING    (URI_MALLOC_ALIGNMENT - sizeof(size_t))
+
+
+
 #define URI_CHECK_ALLOC_OVERFLOW(total_size, nmemb, size) \
 		do { \
 			/* check for unsigned overflow */ \
@@ -170,7 +179,7 @@ void * uriEmulateReallocarray(UriMemoryManager * memory,
 static void * uriDecorateMalloc(UriMemoryManager * memory,
 		size_t size) {
 	UriMemoryManager * backend;
-	const size_t extraBytes = sizeof(size_t);
+	const size_t extraBytes = sizeof(size_t) + URI_MALLOC_PADDING;
 	void * buffer;
 
 	if (memory == NULL) {
@@ -225,7 +234,7 @@ static void * uriDecorateRealloc(UriMemoryManager * memory,
 		return NULL;
 	}
 
-	prevSize = *((size_t *)((char *)ptr - sizeof(size_t)));
+	prevSize = *((size_t *)((char *)ptr - sizeof(size_t) - URI_MALLOC_PADDING));
 
 	/* Anything to do? */
 	if (size <= prevSize) {
@@ -259,7 +268,7 @@ static void uriDecorateFree(UriMemoryManager * memory, void * ptr) {
 		return;
 	}
 
-	backend->free(backend, (char *)ptr - sizeof(size_t));
+	backend->free(backend, (char *)ptr - sizeof(size_t) - URI_MALLOC_PADDING);
 }
 
 
@@ -288,7 +297,7 @@ int uriCompleteMemoryManager(UriMemoryManager * memory,
 
 
 
-int uriTestMemoryManager(UriMemoryManager * memory) {
+int uriTestMemoryManagerEx(UriMemoryManager * memory, UriBool challengeAlignment) {
 	const size_t mallocSize = 7;
 	const size_t callocNmemb = 3;
 	const size_t callocSize = 5;
@@ -456,11 +465,41 @@ int uriTestMemoryManager(UriMemoryManager * memory) {
 		buffer = NULL;
 	}
 
+	/* challenge pointer alignment */
+	if (challengeAlignment == URI_TRUE) {
+		long double * ptr = memory->malloc(memory, 4 * sizeof(long double));
+		if (ptr != NULL) {
+			ptr[0] = 0.0L;
+			ptr[1] = 1.1L;
+			ptr[2] = 2.2L;
+			ptr[3] = 3.3L;
+
+			{
+				long double * const ptrNew = memory->realloc(memory, ptr, 8 * sizeof(long double));
+				if (ptrNew != NULL) {
+					ptr = ptrNew;
+					ptr[4] = 4.4L;
+					ptr[5] = 5.5L;
+					ptr[6] = 6.6L;
+					ptr[7] = 7.7L;
+				}
+			}
+
+			memory->free(memory, ptr);
+		}
+	}
+
 	return URI_SUCCESS;
 }
 
 
 
+int uriTestMemoryManager(UriMemoryManager * memory) {
+    return uriTestMemoryManagerEx(memory, /*challengeAlignment=*/ URI_FALSE);
+}
+
+
+
 /*extern*/ UriMemoryManager defaultMemoryManager = {
 	uriDefaultMalloc,
 	uriDefaultCalloc,