Fix bug #49874: ftell() and fseek() inconsistency when using stream filters

Currently filter seeking does not work correctly for most streams. The
idea is to extend API to allow seeking for some streams. There are
couple of cases:

- filter is always seekable - e.g. string.rot13, string.toupper,
  string.tolower
- filter is seekable only when sought to start or when it re-run the
  data from the start (it means when there is some data buffered) -
  this is for pretty much all other filters that keep some state.
- user filters are seekable by default for BC reason but if new
  seek method is implemented, then it is called and based on the
  bool result the seeking either fails or succeed.

Author: bukka

NEWS

@@ -153,6 +153,8 @@ PHP                                                                        NEWS
   . Fixed bug GH-21221 (Prevent closing of innerstream of php://temp stream).
     (ilutov)
   . Improved stream_socket_server() bind failure error reporting. (ilutov)
+  . Fixed bug #49874 (ftell() and fseek() inconsistency when using stream
+    filters). (Jakub Zelenka)
 
 - Zip:
   . Fixed ZipArchive callback being called after executor has shut down.

UPGRADING.INTERNALS

@@ -74,6 +74,7 @@ PHP 8.6 INTERNALS UPGRADE NOTES
     longer is a pointer, but a directly embedded HashTable struct.
   . Added a C23_ENUM() helper macro to define forward-compatible fixed-size
     enums.
+  . Extended php_stream_filter_ops with seek method.
 
 ========================
 2. Build system changes

ext/bz2/bz2_filter.c

@@ -42,6 +42,10 @@ typedef struct _php_bz2_filter_data {
 	unsigned int is_flushed : 1;          /* only for compression */
 
 	int persistent;
+
+	/* Configuration for reset - immutable */
+	int blockSize100k;  /* compress only */
+	int workFactor;     /* compress only */
 } php_bz2_filter_data;
 
 /* }}} */
@@ -178,6 +182,36 @@ static php_stream_filter_status_t php_bz2_decompress_filter(
 	return exit_status;
 }
 
+static zend_result php_bz2_decompress_seek(
+	php_stream *stream,
+	php_stream_filter *thisfilter,
+	zend_off_t offset,
+	int whence
+	)
+{
+	if (!Z_PTR(thisfilter->abstract)) {
+		return FAILURE;
+	}
+
+	php_bz2_filter_data *data = Z_PTR(thisfilter->abstract);
+
+	/* End current decompression if running */
+	if (data->status == PHP_BZ2_RUNNING) {
+		BZ2_bzDecompressEnd(&(data->strm));
+	}
+
+	/* Reset stream state */
+	data->strm.next_in = data->inbuf;
+	data->strm.avail_in = 0;
+	data->strm.next_out = data->outbuf;
+	data->strm.avail_out = data->outbuf_len;
+	data->status = PHP_BZ2_UNINITIALIZED;
+
+	/* Note: We don't reinitialize here - it will be done on first use in the filter function */
+
+	return SUCCESS;
+}
+
 static void php_bz2_decompress_dtor(php_stream_filter *thisfilter)
 {
 	if (thisfilter && Z_PTR(thisfilter->abstract)) {
@@ -193,6 +227,7 @@ static void php_bz2_decompress_dtor(php_stream_filter *thisfilter)
 
 static const php_stream_filter_ops php_bz2_decompress_ops = {
 	php_bz2_decompress_filter,
+	php_bz2_decompress_seek,
 	php_bz2_decompress_dtor,
 	"bzip2.decompress"
 };
@@ -288,6 +323,41 @@ static php_stream_filter_status_t php_bz2_compress_filter(
 	return exit_status;
 }
 
+static zend_result php_bz2_compress_seek(
+	php_stream *stream,
+	php_stream_filter *thisfilter,
+	zend_off_t offset,
+	int whence
+	)
+{
+	int status;
+
+	if (!Z_PTR(thisfilter->abstract)) {
+		return FAILURE;
+	}
+
+	php_bz2_filter_data *data = Z_PTR(thisfilter->abstract);
+
+	/* End current compression */
+	BZ2_bzCompressEnd(&(data->strm));
+
+	/* Reset stream state */
+	data->strm.next_in = data->inbuf;
+	data->strm.avail_in = 0;
+	data->strm.next_out = data->outbuf;
+	data->strm.avail_out = data->outbuf_len;
+	data->is_flushed = 1;
+
+	/* Reinitialize compression with saved configuration */
+	status = BZ2_bzCompressInit(&(data->strm), data->blockSize100k, 0, data->workFactor);
+	if (status != BZ_OK) {
+		php_error_docref(NULL, E_WARNING, "bzip2.compress: failed to reset compression state");
+		return FAILURE;
+	}
+
+	return SUCCESS;
+}
+
 static void php_bz2_compress_dtor(php_stream_filter *thisfilter)
 {
 	if (Z_PTR(thisfilter->abstract)) {
@@ -301,6 +371,7 @@ static void php_bz2_compress_dtor(php_stream_filter *thisfilter)
 
 static const php_stream_filter_ops php_bz2_compress_ops = {
 	php_bz2_compress_filter,
+	php_bz2_compress_seek,
 	php_bz2_compress_dtor,
 	"bzip2.compress"
 };
@@ -309,7 +380,7 @@ static const php_stream_filter_ops php_bz2_compress_ops = {
 
 /* {{{ bzip2.* common factory */
 
-static php_stream_filter *php_bz2_filter_create(const char *filtername, zval *filterparams, uint8_t persistent)
+static php_stream_filter *php_bz2_filter_create(const char *filtername, zval *filterparams, bool persistent)
 {
 	const php_stream_filter_ops *fops = NULL;
 	php_bz2_filter_data *data;
@@ -388,6 +459,10 @@ static php_stream_filter *php_bz2_filter_create(const char *filtername, zval *fi
 			}
 		}
 
+		/* Save configuration for reset */
+		data->blockSize100k = blockSize100k;
+		data->workFactor = workFactor;
+
 		status = BZ2_bzCompressInit(&(data->strm), blockSize100k, 0, workFactor);
 		data->is_flushed = 1;
 		fops = &php_bz2_compress_ops;
@@ -403,7 +478,7 @@ static php_stream_filter *php_bz2_filter_create(const char *filtername, zval *fi
 		return NULL;
 	}
 
-	return php_stream_filter_alloc(fops, data, persistent);
+	return php_stream_filter_alloc(fops, data, persistent, PSFS_SEEKABLE_START);
 }
 
 const php_stream_filter_factory php_bz2_filter_factory = {

ext/bz2/tests/bz2_filter_seek_compress.phpt

@@ -0,0 +1,55 @@
+--TEST--
+bzip2.compress filter with seek to start
+--EXTENSIONS--
+bz2
+--FILE--
+<?php
+$file = __DIR__ . '/bz2_filter_seek_compress.bz2';
+
+$text1 = 'Short text.';
+$text2 = 'This is a much longer text that will completely overwrite the previous compressed data in the file.';
+
+$fp = fopen($file, 'w+');
+stream_filter_append($fp, 'bzip2.compress', STREAM_FILTER_WRITE);
+
+fwrite($fp, $text1);
+fflush($fp);
+
+$size1 = ftell($fp);
+echo "Size after first write: $size1\n";
+
+$result = fseek($fp, 0, SEEK_SET);
+echo "Seek to start: " . ($result === 0 ? "SUCCESS" : "FAILURE") . "\n";
+
+fwrite($fp, $text2);
+fflush($fp);
+
+$size2 = ftell($fp);
+echo "Size after second write: $size2\n";
+echo "Second write is larger: " . ($size2 > $size1 ? "YES" : "NO") . "\n";
+
+$result = fseek($fp, 50, SEEK_SET);
+echo "Seek to middle: " . ($result === 0 ? "SUCCESS" : "FAILURE") . "\n";
+
+fclose($fp);
+
+$fp = fopen($file, 'r');
+stream_filter_append($fp, 'bzip2.decompress', STREAM_FILTER_READ);
+$content = stream_get_contents($fp);
+fclose($fp);
+
+echo "Decompressed content matches text2: " . ($content === $text2 ? "YES" : "NO") . "\n";
+?>
+--CLEAN--
+<?php
+@unlink(__DIR__ . '/bz2_filter_seek_compress.bz2');
+?>
+--EXPECTF--
+Size after first write: 40
+Seek to start: SUCCESS
+Size after second write: 98
+Second write is larger: YES
+
+Warning: fseek(): Stream filter bzip2.compress is seekable only to start position in %s on line %d
+Seek to middle: FAILURE
+Decompressed content matches text2: YES

ext/bz2/tests/bz2_filter_seek_decompress.phpt

@@ -0,0 +1,43 @@
+--TEST--
+bzip2.decompress filter with seek to start
+--EXTENSIONS--
+bz2
+--FILE--
+<?php
+$file = __DIR__ . '/bz2_filter_seek_decompress.bz2';
+
+$text = 'I am the very model of a modern major general, I\'ve information vegetable, animal, and mineral.';
+
+$fp = fopen($file, 'w');
+stream_filter_append($fp, 'bzip2.compress', STREAM_FILTER_WRITE);
+fwrite($fp, $text);
+fclose($fp);
+
+$fp = fopen($file, 'r');
+stream_filter_append($fp, 'bzip2.decompress', STREAM_FILTER_READ);
+
+$partial = fread($fp, 20);
+echo "First read (20 bytes): " . $partial . "\n";
+
+$result = fseek($fp, 0, SEEK_SET);
+echo "Seek to start: " . ($result === 0 ? "SUCCESS" : "FAILURE") . "\n";
+
+$full = stream_get_contents($fp);
+echo "Content after seek matches: " . ($full === $text ? "YES" : "NO") . "\n";
+
+$result = fseek($fp, 50, SEEK_SET);
+echo "Seek to middle: " . ($result === 0 ? "SUCCESS" : "FAILURE") . "\n";
+
+fclose($fp);
+?>
+--CLEAN--
+<?php
+@unlink(__DIR__ . '/bz2_filter_seek_decompress.bz2');
+?>
+--EXPECTF--
+First read (20 bytes): I am the very model 
+Seek to start: SUCCESS
+Content after seek matches: YES
+
+Warning: fseek(): Stream filter bzip2.decompress is seekable only to start position in %s on line %d
+Seek to middle: FAILURE

ext/gd/tests/bug79945.phpt

@@ -9,17 +9,17 @@ if (!(imagetypes() & IMG_PNG)) {
 }
 set_error_handler(function($errno, $errstr) {
     if (str_contains($errstr, 'Cannot cast a filtered stream on this system')) {
-        die('skip: fopencookie not support on this system');
+        die('skip: fopencookie not supported on this system');
     }
 });
-imagecreatefrompng('php://filter/read=convert.base64-encode/resource=' . __DIR__ . '/test.png');
+imagecreatefrompng('php://filter/read=string.rot13/resource=' . __DIR__ . '/test.png');
 restore_error_handler();
 ?>
 --FILE--
 <?php
-imagecreatefrompng('php://filter/read=convert.base64-encode/resource=' . __DIR__ . '/test.png');
+imagecreatefrompng('php://filter/read=string.rot13/resource=' . __DIR__ . '/test.png');
 ?>
 --CLEAN--
 --EXPECTF--
 
-Warning: imagecreatefrompng(): "php://filter/read=convert.base64-encode/resource=%s" is not a valid PNG file in %s on line %d
+Warning: imagecreatefrompng(): "php://filter/read=string.rot13/resource=%s" is not a valid PNG file in %s on line %d

ext/iconv/iconv.c

@@ -2585,6 +2585,33 @@ static php_stream_filter_status_t php_iconv_stream_filter_do_filter(
 }
 /* }}} */
 
+/* {{{ php_iconv_stream_filter_seek */
+static zend_result php_iconv_stream_filter_seek(
+		php_stream *stream,
+		php_stream_filter *filter,
+		zend_off_t offset,
+		int whence)
+{
+	php_iconv_stream_filter *self = (php_iconv_stream_filter *)Z_PTR(filter->abstract);
+
+	/* Reset stub buffer */
+	self->stub_len = 0;
+
+	/* Reset iconv conversion state by closing and reopening the converter */
+	iconv_close(self->cd);
+
+	self->cd = iconv_open(self->to_charset, self->from_charset);
+	if ((iconv_t)-1 == self->cd) {
+		php_error_docref(NULL, E_WARNING,
+				"iconv stream filter (\"%s\"=>\"%s\"): failed to reset conversion state",
+				self->from_charset, self->to_charset);
+		return FAILURE;
+	}
+
+	return SUCCESS;
+}
+/* }}} */
+
 /* {{{ php_iconv_stream_filter_cleanup */
 static void php_iconv_stream_filter_cleanup(php_stream_filter *filter)
 {
@@ -2595,12 +2622,13 @@ static void php_iconv_stream_filter_cleanup(php_stream_filter *filter)
 
 static const php_stream_filter_ops php_iconv_stream_filter_ops = {
 	php_iconv_stream_filter_do_filter,
+	php_iconv_stream_filter_seek,
 	php_iconv_stream_filter_cleanup,
 	"convert.iconv.*"
 };
 
 /* {{{ php_iconv_stream_filter_create */
-static php_stream_filter *php_iconv_stream_filter_factory_create(const char *name, zval *params, uint8_t persistent)
+static php_stream_filter *php_iconv_stream_filter_factory_create(const char *name, zval *params, bool persistent)
 {
 	php_iconv_stream_filter *inst;
 	const char *from_charset = NULL, *to_charset = NULL;
@@ -2632,7 +2660,8 @@ static php_stream_filter *php_iconv_stream_filter_factory_create(const char *nam
 		return NULL;
 	}
 
-	return php_stream_filter_alloc(&php_iconv_stream_filter_ops, inst, persistent);
+	return php_stream_filter_alloc(&php_iconv_stream_filter_ops, inst, persistent,
+			PSFS_SEEKABLE_START);
 }
 /* }}} */
 

ext/iconv/tests/iconv_stream_filter_seek.phpt

@@ -0,0 +1,43 @@
+--TEST--
+iconv stream filter with seek to start
+--EXTENSIONS--
+iconv
+--FILE--
+<?php
+$file = __DIR__ . '/iconv_stream_filter_seek.txt';
+
+$text = 'Hello, this is a test for iconv stream filter seeking functionality.';
+
+$fp = fopen($file, 'w');
+stream_filter_append($fp, 'convert.iconv.ISO-2022-JP/UTF-8');
+fwrite($fp, $text);
+fclose($fp);
+
+$fp = fopen($file, 'r');
+stream_filter_append($fp, 'convert.iconv.UTF-8/ISO-2022-JP');
+
+$partial = fread($fp, 20);
+echo "First read (20 bytes): " . $partial . "\n";
+
+$result = fseek($fp, 0, SEEK_SET);
+echo "Seek to start: " . ($result === 0 ? "SUCCESS" : "FAILURE") . "\n";
+
+$full = fread($fp, strlen($text));
+echo "Content after seek matches: " . ($full === $text ? "YES" : "NO") . "\n";
+
+$result = fseek($fp, 50, SEEK_SET);
+echo "Seek to middle: " . ($result === 0 ? "SUCCESS" : "FAILURE") . "\n";
+
+fclose($fp);
+?>
+--CLEAN--
+<?php
+@unlink(__DIR__ . '/iconv_stream_filter_seek.txt');
+?>
+--EXPECTF--
+First read (20 bytes): Hello, this is a tes
+Seek to start: SUCCESS
+Content after seek matches: YES
+
+Warning: fseek(): Stream filter convert.iconv.* is seekable only to start position in %s on line %d
+Seek to middle: FAILURE