Bug #73530 [ReO]: Unsetting result set may reset other result set

Edit report at https://bugs.php.net/bug.php?id=73530&edit=1 ID: 73530 Updated by: cmb@php.net Reported by: cmb@php.net Summary: Unsetting result set may reset other result set Status: Re-Opened Type: Bug Package: SQLite related Operating System: * PHP Version: master-Git-2016-11-15 (Git) Assigned To: cmb Block user comment: N Private report: N New Comment: > Please keep in mind that this is also an issue on 5.6.xx and not > only on 7.0.xx/7.1.0! The revert of the BC breaking commit is already in PHP 5.6.30RC1, and as such is supposed to be shipped with PHP 5.6.30. Previous Comments: ------------------------------------------------------------------------ [2017-01-13 10:25:27] nikolay dot petkov at peri dot de Working: PHP 5.6.28 Fails: PHP 5.6.29 I have encountered the same issue with one of our deployed applications that uses the PHPOffice/PHPExcel library to generate Excel part lists. After the PHP has been updated on the server, the application crashes due to the bug fix and the external library. Please keep in mind that this is also an issue on 5.6.xx and not only on 7.0.xx/7.1.0! For now I have to rewrite the application to use another cache method... ------------------------------------------------------------------------ [2016-12-29 15:19:38] jcrews at gridlox dot net >Thanks for reporting! > >> It appears, via static analysis, the correct solution is to >> sqlite3_reset and sqlite3_clear prior to re-binding parameters >> in sqlite3stmt::execute (in addition to this change). > >That doesn't appear to solve the issue with your supplied test >script. I'm not yet sure why, but probably it hasn't been a good >idea to try to resolve this issue in a patch release anyway, so >I'm going to revert the fix for now. I will be a good OSS citizen and work on a solution for 7.1.x. Thank you for reverting in the interim. ------------------------------------------------------------------------ [2016-12-29 13:03:01] cmb@php.net Automatic comment on behalf of cmbecker69@gmx.de Revision: http://git.php.net/?p=php-src.git;a=commit;h=2ba3b275948050ce600c5234b66e840b640ca5a5 Log: Revert &quot;Fix #73530: Unsetting result set may reset other result set&quot; ------------------------------------------------------------------------ [2016-12-29 11:56:25] cmb@php.net > The change referenced in this bug introduced a regression > compared to earlier releases. Thanks for reporting! > It appears, via static analysis, the correct solution is to > sqlite3_reset and sqlite3_clear prior to re-binding parameters > in sqlite3stmt::execute (in addition to this change). That doesn't appear to solve the issue with your supplied test script. I'm not yet sure why, but probably it hasn't been a good idea to try to resolve this issue in a patch release anyway, so I'm going to revert the fix for now. ------------------------------------------------------------------------ [2016-12-29 03:40:29] jcrews at gridlox dot net The change referenced in this bug introduced a regression compared to earlier releases. Working: php-7.0.13 Fails: php-7.0.14 Description ---------------- Using php-7.0.14 with phpoffice/phpexcel with sqlite3 caching causes documents to fail to load. This is ultimately caused by phpexcel not using SQLite3Result::finalize (or reset) internally, within limited-lifetime functions. This is a significant change in behavior that may break deployed applications during system maintenance. sqlite3_step has two effects: - On a reset (or new) statement, the prepared statement will be executed, and the first row will be available. - On a statement VM that previously had sqlite3_step called, the next result row will be available, until the end of data is reached. sqlite3_reset resets the statement for execution, not any bound parameters. This also explains the original test script behavior: A step is peformed after a reset, which was implicitly called through unset. The statement was thus re-run, returning the first result, again. When fetchArray hits SQLITE_DONE, the statement isn't reset, and it shouldn't be. Otherwise, any while(haveData) loops would run forever. It appears, via static analysis, the correct solution is to sqlite3_reset and sqlite3_clear prior to re-binding parameters in sqlite3stmt::execute (in addition to this change). This will guarantee that each call to execute provides consistent results. Unfortunately, I do not have sufficient resources to build and test php at this time. Test Script ---------------- <?php $db = new SQLite3(':memory:'); $db->exec("CREATE TABLE foo (num int)"); $db->exec("INSERT INTO foo VALUES (0)"); $db->exec("INSERT INTO foo VALUES (1)"); $stmt = $db->prepare("SELECT * FROM foo WHERE NUM = ?"); function pass1($db, $stmt) { $stmt->bindValue(1, 0, SQLITE3_INTEGER); $res1 = $stmt->execute(); while ($row = $res1->fetchArray(SQLITE3_ASSOC)) { var_dump($row); } echo "finished first pass\n"; } function pass2($db, $stmt) { $stmt->bindValue(1, 1, SQLITE3_INTEGER); $res2 = $stmt->execute(); /* Oops, we forgot to reset in fetchArray, no effect */ while ($row = $res2->fetchArray(SQLITE3_ASSOC)) { var_dump($row); } } pass1($db, $stmt); pass2($db, $stmt); ?> Expected Result (>>php-7.0.13): ---------------- sqlitetest.php:12: array(1) { 'num' => int(0) } finished first pass sqlitetest.php:21: array(1) { 'num' => int(1) } Observed Result: ---------------- sqlitetest.php:12: array(1) { 'num' => int(0) } finished first pass sqlitetest.php:21: array(1) { 'num' => int(0) } ------------------------------------------------------------------------ The remainder of the comments for this report are too long. To view the rest of the comments, please view the bug report online at https://bugs.php.net/bug.php?id=73530 -- Edit this bug report at https://bugs.php.net/bug.php?id=73530&edit=1