I have gone somewhat crazy troubleshooting this problem, but I think I have something that may possibly be useful to you.

I've been getting a ton of segmentation faults (blank white screen in browser) in apache, and it seemed to be happening only with drupal sites I had created on subdomains. Took me forever, but this is what seems to be reliably triggering the segfaults:

  • Website established on a subdomain (ie clients.domain.com where www.domain.com already exists)
  • User is not user ID 1
  • Administration menu module enabled and "access administration menu" permission in effect for user
  • Comment module enabled and "administer comments" permission in effect
  • Spam module enabled and "administer spam" permission in effect
  • "access administration pages," "access user profiles," and "administer users" permissions also in effect, but I think these are irrelevant to the problem itself
  • User visits a user profile, ie http://clients.domain.com/user/22 - segfault does not happen every time this page is loaded, but a significant minority (20 - 40% of page loads)

By now you think I'm insane, I'm sure. I don't know if you'll be able to reproduce this problem, but it's certainly the case on my machine. I guess I'm posting this on the off chance that someone else ends up googling "site:drupal.org segmentation fault" and has the same problem.

Versions:

MediaTemple (dv) server using PLESK to setup domains/subdomains, running CentOS
Apache 2
Drupal 5.2
admin_menu.info,v 1.3 2007/02/10 18:18:34
spam.module 5.x-1.0, datestamp = "1181294108"
comment.info,v 1.3 2006/11/21 20:55:34 dries

Comments

greenmachine’s picture

greenmachine commented

Oh yeah, obviously this is a server issue so it is browser-independent. Examples from the central apache log (/var/log/httpd/error_log):

[Wed Oct 10 12:23:11 2007] [notice] child pid 32210 exit signal Segmentation fault (11)
[Wed Oct 10 12:23:44 2007] [notice] child pid 11938 exit signal Segmentation fault (11)
[Wed Oct 10 12:23:52 2007] [notice] child pid 13528 exit signal Segmentation fault (11)
[Wed Oct 10 12:25:49 2007] [notice] child pid 16170 exit signal Segmentation fault (11)

Each segfault coincided with a blank page encountered in the browser.

sun’s picture

sun
Primary language German
Location Karlsruhe
commented

Now, *that* sounds strange... I will definitely try to reproduce this ASAP. However, since I have to complete a bunch of projects currently, I cannot guarantee when this will happen.

Could you also supply your PHP version, please?

rcross’s picture

rcross commented

Has this been reproduced? I seem to be getting a similar problem that's been driving me up the wall. It seems to mainly be targeting comments (which makes my forums unusable).

sun’s picture

sun
Primary language German
Location Karlsruhe
commented

@greenmachine: I have tried to reproduce the segmentation faults in your mentioned setup now, but with no "luck". This probably means that we are dealing with a platform specific bug. Some ideas:

  • Update to the latest versions of admin_menu and spam module (if not yet done)
  • Check if the segmentation faults also occur with all RTBC patches in the current queue
  • Check if the bug also occurs after applying this patch
  • Try to set PHP error reporting to E^ALL and look if there are any messages

@rcroos: You need to supply further information about your setup. Take greenmachine's excellent write-up in this issue as an example.

sun’s picture

sun
Primary language German
Location Karlsruhe
commented
Status: Active » Postponed (maintainer needs more info)

FYI: Almost all PNR and RTBC patches have been committed now. A new development snapshot will be available in approx. 12 hours.

rcross’s picture

rcross commented

Not sure if this helps, but I realized what was causing my issue was the op-code cache/php accelerator that was installed on my server. Please note the drawbacks section on this site http://2bits.com/articles/php-op-code-caches-accelerators-a-must-for-a-l...

sun’s picture

sun
Primary language German
Location Karlsruhe
commented

@greenmachine: Any updates? I would like to see this issue resolved, because it might scare some users.

sun’s picture

sun
Primary language German
Location Karlsruhe
commented
Status: Postponed (maintainer needs more info) » Closed (works as designed)

Since no further feedback was provided and it's probably related to an opcode cache, I'm obliged to mark this 'by design'.

Feel free to re-open this issue if you want to supply further information.