ID: 27231 Updated by: [EMAIL PROTECTED] Reported By: herman at frontier dot nl -Status: Open +Status: Verified Bug Type: iPlanet related Operating System: SunOS 5.8+ PHP Version: 4CVS-2004-02-13
Previous Comments: ------------------------------------------------------------------------ [2004-02-13 18:32:00] [EMAIL PROTECTED] You do not need to test ist on your system. I can reproduce the crashs here, even with the latest stable snapshot. I think we should try to reproduce this with other multithreaded servers to check if it is a ZTS bug (I think it is one). ------------------------------------------------------------------------ [2004-02-13 09:00:58] [EMAIL PROTECTED] Same on PHP 4.3.5 RC2 of SunOS 5.9. Tried to debug it but the crashing process did not create a core dump. If somebody of the others helps this: # /opt/forte7/SUNWspro/bin/dbx For information about new features see `help changes' To remove this message, put `dbxenv suppress_startup_message 7.0' in your .dbxrc (dbx) attach 9702 ... ... detected a multithreaded program Attached to process 9702 with 90 LWPs [EMAIL PROTECTED] ([EMAIL PROTECTED]) stopped in __lwp_park at 0xfe3e5f88 0xfe3e5f88: __lwp_park+0x0010: ta %icc,%g0 + 8 (dbx) cont -> here starting of crashing test2.php [EMAIL PROTECTED] ([EMAIL PROTECTED]) signal SEGV (no mapping at the fault address) in zend_clean_garbage at line 25 in file "zend_execute_locks.h" 25 while (EG(garbage_ptr)) { dbx: read of 4 bytes at address ee2cf748 failed -- Error 0 (dbx) where current thread: [EMAIL PROTECTED] =>[1] zend_clean_garbage(tsrm_ls = <bad address 0xee2cf7dc>), line 25 in "zend_execute_locks.h" dbx: read of 4 bytes at address ee2cf7b8 failed -- Error 0 dbx: attempt to read frame failed -- cannot derive frame pointer (dbx) seems to be a TSRM problem because in CLI it does not appear. And crash is not in NSAPI code. ------------------------------------------------------------------------ [2004-02-13 03:42:22] herman at frontier dot nl Sorry, forgot to mention: no, using the CLI (at least scripts #1 and #2) they run OK. ------------------------------------------------------------------------ [2004-02-12 20:52:26] [EMAIL PROTECTED] Does it crash with PHP-CLI? ------------------------------------------------------------------------ [2004-02-12 14:59:48] herman at frontier dot nl Description: ------------ When trying an PHP application (which uses generated code, some rather quite large files) on Sun server running Solaris 8, the SunONE webserver 6.1 and PHP 4.3.4 using nsapi, one of the scripts crashed the server. After some experimenting, it seems the size of the script file triggered the crash: a PHP script > 128KB will bring down the server, a script < 128KB will not. To complicate matters a bit, if a small script includes a large script, the server will also crash, but if a small script will include two "half size" scripts, it will not. The server seems to run other PHP scripts well, and is also used to run applets through its nsapi interface without problems. PHP has been compiled as per instructions on the PHP website, with mysql disabled, and ldap and oci8 enabled. Reproduce code: --------------- The test scripts consist of: <?php echo "hello"; exit; $a = "123456789012345678901234567890123456789012345678901234567890"; // previous line copied till the script is large enough ?> The scripts can be found at: http://www.ozuzo.net/phpbug/test1.php.txt => WORKS http://www.ozuzo.net/phpbug/test2.php.txt => CRASH Include examples: http://www.ozuzo.net/phpbug/test3.php.txt => WORKS http://www.ozuzo.net/phpbug/test3a.php.txt http://www.ozuzo.net/phpbug/test3b.php.txt http://www.ozuzo.net/phpbug/test4.php.txt => CRASH http://www.ozuzo.net/phpbug/test4a.php.txt Expected result: ---------------- The word "hello". Actual result: -------------- The server crashes with the following message (no entry in the PHP log): failure (25268): CORE3107: Child process closed admin channel fine (25268): CORE3061: signal_handler_thread: received signal 18 fine (25268): CORE3049: Primordial process detected child 25296 died: status 11 fine (25268): CORE3050: Is our child, will spawn replacement fine (25268): CORE3062: Unlinking of /tmp/<servername>/.cgistub_25296 returned -1 fine (25268): CORE3047: Server spawned worker process 25305 ------------------------------------------------------------------------ -- Edit this bug report at http://bugs.php.net/?id=27231&edit=1