Re: Re: Test 4784 (5.2.10rc2) hung my server

Andy Shellam wrote:

Hi, I've just run "make test" again to make sure it wasn't a one-off, and unfortunately it wasn't. Again, the final output from the command (I've included the actual test that's causing the problem too.) I've attached another screenshot showing the difference in processor usage when the test actually started. Note the previous tests don't appear to hit more than 50-60% usage. The drop-off from 100% to near-zero is when I cancelled it. PASS Bug #26003 (fgetcsv() is not binary-safe) [ext/standard/tests/file/bug26003.phpt] PASS Bug #26615 (exec crash on long input lines) [ext/standard/tests/file/bug26615.phpt] PASS Bug #26938 (exec does not read consecutive long lines correctly) [ext/standard/tests/file/bug26938.phpt] TEST 4784/7161 [ext/standard/tests/file/bug27508.phpt] Hope this helps. (PS. Please copy me into replies as I'm not registered on the list.)

Andy, It sounds like an infinite loop. If the problem persists, log a bug at http://bugs.php.net/ Check that it occurs with the final 5.2.10 release. Identify the failing script (I presume it is ext/standard/tests/file/bug27508.phpt. If not, look for the next file in that directory). Some other useful info would be a system call trace (truss or strace) on the running process. Try building PHP with --enable-debug and run the PHP CLI (in php-5.2.10/sapi/cli/php) in a debugger. Issue a break and see where the program stops. Also, send any email follow-ups to the internals@lists.php.net list instead of php-qa. Chris

Thanks, Andy Andy Shellam wrote:

Hi all, I'm just posting to report an issue with test 4784 of PHP 5.2.10rc2. I've compiled the above PHP version on my development server, and when it hit this test, it locked my system up. My development server is running Solaris 10 5/09 in a VMware Workstation 6.5 virtual machine running on Windows Server 2003. The physical server has 2GB RAM, of which 1GB is dedicated to the VM. Shortly after the test started, the machine's processor load hit 100% and stuck on that for about 10 minutes; it then averaged 80-90% until I cancelled the "make test." While it was stuck at 100%, the machine wouldn't do anything - logging in, running any commands etc - when I eventually managed to get another SSH connection to it, every command returned a "fork: resource unavailable" message. Please see the attached screenshot showing the processor usage. The last output before I had to cancel is below: PASS Bug #22414 (passthru() does not read data correctly) [ext/standard/tests/file/bug22414.phpt] PASS Bug #24313 (file_exists() throws a warning on nonexistent files when is open_basedir enabled) [ext/standard/tests/file/bug24313.phpt] PASS Bug #24482 (GLOB_ONLYDIR not working) [ext/standard/tests/file/bug24482.phpt] PASS Bug #26003 (fgetcsv() is not binary-safe) [ext/standard/tests/file/bug26003.phpt] PASS Bug #26615 (exec crash on long input lines) [ext/standard/tests/file/bug26615.phpt] PASS Bug #26938 (exec does not read consecutive long lines correctly) [ext/standard/tests/file/bug26938.phpt] ^C^C^C^C^C^C^C And here's the output from the start of the test sequence: =====================================================================

PHP         : /var/software/php-5.2.10RC2/sapi/cli/php
PHP_SAPI    : cli

PHP_VERSION : 5.2.10RC2 ZEND_VERSION: 2.2.0

PHP_OS      : SunOS - SunOS dev1.networkmail.eu 5.10 Generic_139556-08

i86pc INI actual : /var/software/php-5.2.10RC2/tmp-php.ini

More .INIs  : CWD         : /var/software/php-5.2.10RC2

Extra dirs :

VALGRIND    : Not used

===================================================================== I've never had any previous issues running the PHP tests - 5.2.9 compiled and ran all tests fine. Thanks, Andy ------------------------------------------------------------------------

-- Email: christopher.jones@oracle.com Twitter: http://twitter.com/ghrd