ID: 14365 Comment by: [EMAIL PROTECTED] Reported By: [EMAIL PROTECTED] Status: Feedback Bug Type: iPlanet related Operating System: RedHat Linux 7.3 PHP Version: 4.3.0-dev New Comment:
Wow! I managed to create a page that creates a segfault every time... Before PHP only crashed once in a while (once in 40 requests?) but now I finally have something reproducable. The only thing I did was adding an extra INSERT query to the end of a search.php script. Haven't tried to create a minimal testcase yet, just to glad tat I can finally do some more testing. BTW IMO it's a good idea to change this bug. It's not about require_once() but about canonicalize() crashing PHP 4.3.0 (and something on 4.2.3 that I haven't been able to pinpoint yet). And it's not limited to iPlanet servers. Are there any more ways to get more information out of PHP? Memory usage? Most used functions? That kind of information? I'm going to play a bit with standalone httpd's so I don't have to attach to a running server. Oh and is it possible to subscribe to a bug at bugs.php.net? Previous Comments: ------------------------------------------------------------------------ [2003-01-06 06:17:35] [EMAIL PROTECTED] We are using evals but the worst I could find was 65 evals on one page (the template system uses evals for nested blocks). The index page has a complex template and needs 55 evals, normal pages only need 4 or 5 evals. ------------------------------------------------------------------------ [2003-01-06 01:58:31] [EMAIL PROTECTED] I've read the backtrace page but the bt there doesn't show php filenames and linenumber.. But Sean mailed me that I should try 'bt full', I'll try that the next time. Anyway, as I said I'm using Apache and not iPlanet... The canonicalize problem only seems to occur on 4.3.0, PHP 4.2.3 seems to segfault on memcopy() ------------------------------------------------------------------------ [2003-01-05 09:33:15] [EMAIL PROTECTED] See http://bugs.php.net/bugs-generating-backtrace.php for details on making backtraces. The stack size appears to be a setting in the iPlanet configuration:Try modifying your ns/threads section to include a larger stacksize...the default one is 128*1024 I ran into a stacksize problem on the recent redhat update (related to glibc) that was segfaulting on a dns lookup. I didn't associate it with the segfault I was getting for PHP (Including squirrelmail), but the segfault for SM is gone now for me after this fix and SM works perfectly. Try this sample: ns_section "ns/threads" ns_param stacksize [expr 256*1024] " (quoted from this bug report). I don't have and have never used iPlanet, so it's up to you to find out how to configure it. " ------------------------------------------------------------------------ [2003-01-05 09:07:05] [EMAIL PROTECTED] I would love to but I don't have a clue how I should do that :D I'm using Apache 1.3.26 and PHP 4.3.0 (but that one segfaults too often) and PHP 4.2.3 (segfaults too but less often, especially when we use a debug build and limit the MaxRequestsPerChild to 10. Nested evals()? Hmm, we shouldn't, I'll go and see if that can be the case. BTW how did Sean create that backtrace? It even shows in what PHP file and on what line Apache/PHP went down... ------------------------------------------------------------------------ [2003-01-02 11:54:51] [EMAIL PROTECTED] Can you try increasing your stack size as was suggested in one of the earlier responses? The latest backtrace looks *very* deep - are you using a lot of nested eval() calls ? ------------------------------------------------------------------------ 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 http://bugs.php.net/14365 -- Edit this bug report at http://bugs.php.net/?id=14365&edit=1
