Bug #79811 [Opn]: Script stops and delivers incomplete web page

Edit report at https://bugs.php.net/bug.php?id=79811&edit=1 ID: 79811 Updated by: cmb@php.net Reported by: buschmann at nidsa dot net Summary: Script stops and delivers incomplete web page Status: Open Type: Bug Package: opcache Operating System: Windows Server 2019 64bit PHP Version: 8.0.0alpha1 Block user comment: N Private report: N New Comment: The point is that pg_fetch_assoc($rSOrder) is called while $rSOrder is undefined. This is supposed to throw a TypeError in PHP 8. This TypeError is apparently caught by your script (maybe by set_error_handler() or set_exception_handler()). At least for debugging purposes, you could rewrite @$fSOrder=pg_fetch_assoc($rSOrder); to $fsOrder = isset($rSOrder) ? pg_fetch_assoc($rSOrder) : []; or so. Previous Comments: ------------------------------------------------------------------------ [2020-07-09 07:09:30] buschmann at nidsa dot net I have tried it with opcache disabled, same error: The Zend OPcache extension is installed but not active The code comes from the php 4 era and has no explicit exception handling. The variable $rSOrder is only declared inside of this function as global and not set before. It should be easyly reproducable by declaring the function, calling the loop with dummy echos and some dummy echos after the loop. The core problem is that the script stops processing (different from 7.x.y). ------------------------------------------------------------------------ [2020-07-08 16:46:56] cmb@php.net > The call is probably throwing a TypeError, but something in your > error/exception handling setting eats it. That's quite likely. However, to exclude any potential OPcache issues, please run the script with OPcache disabled. ------------------------------------------------------------------------ [2020-07-08 16:27:10] nikic@php.net The call is probably throwing a TypeError, but something in your error/exception handling setting eats it. ------------------------------------------------------------------------ [2020-07-08 16:23:10] buschmann at nidsa dot net Description: ------------ During testing our application for compatibility with php 8 one script stopped suddenly and the resulting web page was incomplete. This occured when loading the empty web page without retrieving the central data from the database. The stop occurs at the following lines of code (the original code): ... for($i=0;$i<=7;$i++){ @$fSOrder=pg_fetch_assoc($rSOrder); if($fSOrder['od_distribution']==2){ ... This code has worked on all the version 7.x.y until 7.4.8 even if it misses error-handling. The query to get a valid $rSOrder with pg_query has not been called/executed, so $rSOrder is not existent at this point of the script. ($rSOrder etc. is declared global in a function for executing the query: function searchOrder(){ global $fSOrder,$rSOrder,$book,$iden,$style,$season,$file,$only; ... The original application developers knew that at this point there could be an error by using the not existing $rSOrder and included the @ to suppress the error message (not introduced/changed by me). When using php8 (already tested on php8 alpha 2) the script terminates with a last error message pointing to the line of pg_fetch_assoc: 08-Jul-2020 16:35:20 Europe/Berlin] PHP Warning: Undefined variable $rSOrder in N:\cpsdev\cps\cps_pg\AltaPedido.php on line 2390 The script terminates processing without any fatal error/crash and sends the incomplete web page to apache/browser. On the same machine and the same source tree php 7.4.8 logs the same error but continues and completes the web page. The problem remains when removing error reporting by deleting the @. Environment: - Windows server 2019 64bit Standard in a VM under HyperV - Apache 2.43 64bit from apachelounge - php 8 alpha2 64bit thread safe from php.net QA download - php loaded as a module with php_module - JIT is on for php8 (not sure if it is really used) - opcache is enabled ------------------------------------------------------------------------ -- Edit this bug report at https://bugs.php.net/bug.php?id=79811&edit=1