#18165 closed defect (bug) (worksforme)
Admin tables empty when installed on *nix with Apache 2.x, PHP 5.3 running as CGI with wrapper script
| Reported by: |
|
Owned by: | |
|---|---|---|---|
| Milestone: | Priority: | normal | |
| Severity: | normal | Version: | 3.2.1 |
| Component: | Administration | Keywords: | has-patch |
| Focuses: | Cc: |
Description
- Wordpress 3.2.1
- Apache 2.2.3
- PHP 5.3.5
When running with suexec through a wrapper script such as set up by Webmin, the PHP_SELF variable is "fixed" by PHP to point to the wrapper instead of the actual script. This in turn kills some of the admin lists (posts are empty) along with problems setting theme (Thematic) options.
To recreate, adjust httpd.conf vhost to:
<VirtualHost *:80>
SuexecUserGroup "#1111" "#2222"
ServerName example.com
DocumentRoot /home/example.com/www
ErrorLog /home/example.com/logs/error_log
CustomLog /home/example.com/logs/access_log combined
ScriptAlias /cgi-bin/ /home/example.com/cgi-bin/
<Directory /home/example.com/www>
Options Indexes IncludesNOEXEC FollowSymLinks
allow from all
AllowOverride All
Action application/x-httpd-php5 /cgi-bin/php5.cgi
AddType application/x-httpd-php5 .php5
AddType application/x-httpd-php5 .php
</Directory>
<Directory /home/example.com/cgi-bin>
allow from all
</Directory>
RemoveHandler .php
RemoveHandler .php5
</VirtualHost>
Add the following shell script /cgi-bin/php5.cgi
#!/bin/bash PHPRC=$DOCUMENT_ROOT/../etc/php.ini export PHPRC umask 022 SCRIPT_FILENAME=$PATH_TRANSLATED export SCRIPT_FILENAME exec /usr/bin/php-cgi
Copy the php.ini file to /home/example.com/etc/
The problem is temporarily solved by setting the following PHP option in php.ini:
cgi.fix_pathinfo = 0
Which tell PHP not to do a path gawk. I think what should be done is Wordpress should not use PHP_SELF or at least prefer using ORIG_PHP_SELF if it exists, possibly resetting PHP_SELF for downstream functions.
Attachments (2)
Change History (9)
#2
@
14 years ago
- Keywords has-patch added
@magdude, I managed to get my setup to match yours and was able to reproduce this. I've attached 18165.patch, which has resolved things on my end. Please report back with your test results.
I'm not able to recreate this with 3.3-trunk using the configuration you've provided. Can you recreate this with 3.3? If not, can you provide more information about your setup, perhaps screen shots / logs / etc about what's breaking?