Fix bug #52335 (fseek() on memory stream behavior different then file)

This changes memory stream to allow seeking past end which makes it the
same as seeking on files. It means the position is allowed to be higher
than the string length. The size only increases if data is appended to
the past position. The space between the previous string and position
is filled with zero bytes.

Fixes GH-9441
Closes GH-12058

Author: bukka

NEWS

@@ -20,6 +20,10 @@ PHP                                                                        NEWS
   . Fixed GH-11982 (str_getcsv returns null byte for unterminated enclosure).
     (Jakub Zelenka)
 
+- Streams:
+  . Fixed bug #52335 (fseek() on memory stream behavior different than file).
+    (Jakub Zelenka)
+
 17 Aug 2023, PHP 8.3.0beta3
 
 - Core:

UPGRADING

@@ -648,6 +648,9 @@ PHP 8.3 UPGRADE NOTES
 - Streams:
   . Blocking fread() on socket connection returns immediately if there are
     any buffered data instead of waiting for more data.
+  . Memory stream no longer fails if seek offset is past the end. Instead
+    the memory is increase on the next write and date between the old end and
+    offset is filled with zero bytes in the same way how it works for files.
 
 ========================================
 14. Performance Improvements

ext/standard/tests/file/bug52335.phpt

@@ -0,0 +1,67 @@
+--TEST--
+Bug #52335 (fseek() on memory stream behavior different then file)
+--FILE--
+<?php
+
+echo "Read mode\n";
+$fpr = fopen("php://memory", "r");
+var_dump(fseek($fpr, 20));
+var_dump(feof($fpr));
+var_dump(ftell($fpr));
+var_dump(feof($fpr));
+var_dump(fread($fpr, 2));
+var_dump(feof($fpr));
+var_dump(fseek($fpr, 24));
+var_dump(feof($fpr));
+var_dump(ftell($fpr));
+fclose($fpr);
+
+echo "Read write mode\n";
+$fprw = fopen("php://memory", "r+");
+var_dump(fwrite($fprw, "data"));
+var_dump(fseek($fprw, 20, SEEK_END));
+var_dump(feof($fprw));
+var_dump(ftell($fprw));
+var_dump(feof($fprw));
+var_dump(fread($fprw, 2));
+var_dump(feof($fprw));
+var_dump(fseek($fprw, 20));
+var_dump(fwrite($fprw, " and more data"));
+var_dump(feof($fprw));
+var_dump(ftell($fprw));
+var_dump(fread($fprw, 10));
+var_dump(fseek($fprw, 16, SEEK_CUR));
+var_dump(ftell($fprw));
+var_dump(fseek($fprw, 0));
+var_dump(bin2hex(stream_get_contents($fprw)));
+fclose($fprw);
+
+?>
+--EXPECT--
+Read mode
+int(0)
+bool(false)
+int(20)
+bool(false)
+string(0) ""
+bool(true)
+int(0)
+bool(false)
+int(24)
+Read write mode
+int(4)
+int(0)
+bool(false)
+int(24)
+bool(false)
+string(0) ""
+bool(true)
+int(0)
+int(14)
+bool(false)
+int(34)
+string(0) ""
+int(0)
+int(50)
+int(0)
+string(68) "646174610000000000000000000000000000000020616e64206d6f72652064617461"

ext/standard/tests/file/gh9441.phpt

@@ -0,0 +1,20 @@
+--TEST--
+Bug GH-9441 (fseek does not work with php://input when data is not preread)
+--INI--
+enable_post_data_reading=0
+--POST_RAW--
+Content-Type: application/unknown
+a=123456789&b=ZYX
+--FILE--
+<?php
+$input = fopen("php://input", "r");
+var_dump(fseek($input, 10));
+var_dump(ftell($input));
+var_dump(fread($input, 10));
+var_dump(file_get_contents("php://input"));
+?>
+--EXPECT--
+int(0)
+int(10)
+string(7) "9&b=ZYX"
+string(17) "a=123456789&b=ZYX"

ext/standard/tests/file/stream_rfc2397_007.phpt

@@ -125,8 +125,8 @@ int(0)
 int(3)
 bool(false)
 ===S:10,C===
-int(-1)
-bool(false)
+int(0)
+int(13)
 bool(false)
 ===S:-1,E===
 int(0)
@@ -137,6 +137,6 @@ int(0)
 int(6)
 bool(false)
 ===S:1,E===
-int(-1)
-bool(false)
+int(0)
+int(7)
 bool(false)

main/streams/memory.c

@@ -49,11 +49,17 @@ static ssize_t php_stream_memory_write(php_stream *stream, const char *buf, size
 
 	if (ms->mode & TEMP_STREAM_READONLY) {
 		return (ssize_t) -1;
-	} else if (ms->mode & TEMP_STREAM_APPEND) {
-		ms->fpos = ZSTR_LEN(ms->data);
 	}
-	if (ms->fpos + count > ZSTR_LEN(ms->data)) {
+	size_t data_len = ZSTR_LEN(ms->data);
+	if (ms->mode & TEMP_STREAM_APPEND) {
+		ms->fpos = data_len;
+	}
+	if (ms->fpos + count > data_len) {
 		ms->data = zend_string_realloc(ms->data, ms->fpos + count, 0);
+		if (ms->fpos > data_len) {
+			/* zero the bytes added due to seek past end position */
+			memset(ZSTR_VAL(ms->data) + data_len, 0, ms->fpos - data_len);
+		}
 	} else {
 		ms->data = zend_string_separate(ms->data, 0);
 	}
@@ -73,7 +79,7 @@ static ssize_t php_stream_memory_read(php_stream *stream, char *buf, size_t coun
 	php_stream_memory_data *ms = (php_stream_memory_data*)stream->abstract;
 	assert(ms != NULL);
 
-	if (ms->fpos == ZSTR_LEN(ms->data)) {
+	if (ms->fpos >= ZSTR_LEN(ms->data)) {
 		stream->eof = 1;
 		count = 0;
 	} else {
@@ -132,20 +138,14 @@ static int php_stream_memory_seek(php_stream *stream, zend_off_t offset, int whe
 					return 0;
 				}
 			} else {
-				if (ms->fpos + (size_t)(offset) > ZSTR_LEN(ms->data)) {
-					ms->fpos = ZSTR_LEN(ms->data);
-					*newoffs = -1;
-					return -1;
-				} else {
-					ms->fpos = ms->fpos + offset;
-					*newoffs = ms->fpos;
-					stream->eof = 0;
-					return 0;
-				}
+				stream->eof = 0;
+				ms->fpos = ms->fpos + offset;
+				*newoffs = ms->fpos;
+				return 0;
 			}
 		case SEEK_SET:
-			if (ZSTR_LEN(ms->data) < (size_t)(offset)) {
-				ms->fpos = ZSTR_LEN(ms->data);
+			if (offset < 0) {
+				ms->fpos = 0;
 				*newoffs = -1;
 				return -1;
 			} else {
@@ -156,9 +156,10 @@ static int php_stream_memory_seek(php_stream *stream, zend_off_t offset, int whe
 			}
 		case SEEK_END:
 			if (offset > 0) {
-				ms->fpos = ZSTR_LEN(ms->data);
-				*newoffs = -1;
-				return -1;
+				ms->fpos = ZSTR_LEN(ms->data) + offset;
+				*newoffs = ms->fpos;
+				stream->eof = 0;
+				return 0;
 			} else if (ZSTR_LEN(ms->data) < (size_t)(-offset)) {
 				ms->fpos = 0;
 				*newoffs = -1;