If php.ini is incorrect, php-frm starts without warning with default values

[selivan]

OS: Ubuntu 16.04 Xenial
PHP: php7.0-fpm 7.0.8-0ubuntu0.16.04.2

If /etc/php/7.0/fpm/php.ini has syntax error, php7.0-fpm starts silently (!! no error messages even in logs), but uses default values.

To make php.ini incorrect, just add this line:

# Wrong comment (

Systemd servive unit for php7.0-fpm has config check:

ExecStartPre=/usr/lib/php/php7.0-fpm-checkconf

But is does not work:

root@xenial:~# /usr/sbin/php-fpm7.0 --fpm-config /etc/php/7.0/fpm/php-fpm.conf --test
PHP: syntax error, unexpected '(' in /etc/php/7.0/fpm/php.ini on line 6
[14-Sep-2016 14:24:46] NOTICE: configuration file /etc/php/7.0/fpm/php-fpm.conf test is successful

root@xenial:~# /usr/lib/php/php7.0-fpm-checkconf; echo $?
0

So, if php.ini is incorrect, php-fpm silently starts with default values: post_max_size=8m, expose_php is enabled, disable_functions is empty and so on.

[Nish Aravamudan (nacc)]

Hello and thank you for reporting this bug.

With 16.10, the helper script has been removed, so the problem is not exactly reproducible. With 16.04, it looks to be a simple typo with the grep ([ERROR] rather than ERROR) in the script. Modifying it in that way leads to a successful prevention of php-fpm starting on parse errors. I will work on fixing 16.10 (if possible) and then SRU'ing the fix for 16.04.

Thanks,
Nish

[Nish Aravamudan (nacc)]

16.10's dropping of the helper script, I believe, perhaps in coordination with other changes, means that if you make a similar modification there and try to restart the service you get:

# service php7.0-fpm restart
Job for php7.0-fpm.service failed because the control process exited with error code.
See "systemctl status php7.0-fpm.service" and "journalctl -xe" for details.

So it's already fixed in 16.10. I'll look at what's needed to make the same occur in 16.04.

[Nish Aravamudan (nacc)]

While the helper script can/should be fixed, I'm actually unable to reproduce this on 16.04:

# vi /etc/php/7.0/fpm/php-fpm.conf
[insert: "# wrong comment (" at the top of the file]
# service php7.0-fpm restart
Job for php7.0-fpm.service failed because the control process exited with error code. See "systemctl status php7.0-fpm.service" and "journalctl -xe" for details.

Can you give more details on how you installed php-fpm? Was this an upgrade from 14.04? Or did you install 16.04 at release and follow the normal upgrade paths to 16.04.1?

[selivan (selivan5)]

> # vi /etc/php/7.0/fpm/php-fpm.conf

no, you should do

# vi /etc/php/7.0/fpm/php.ini

It handles errors in php-fpm.conf, but silently ignores in php.ini

[selivan (selivan5)]

First I saw this bug on trusty with php7 ppa, where I upgraded from 5.5 by copying php.ini. It had inside comment like "# blablabla (see foobar)".

I created bug report for ppa owner: https://github.com/oerdnj/deb.sury.org/issues/456

Then I reproduced it out on newly created xenial instance in AWS and reported it here.

[Nish Aravamudan (nacc)]

Sorry for my mistake! I have corrected the tasks states -- and reproduce the issue on 16.10.

[selivan (selivan5)]

Also reported it directly to php guys: https://bugs.php.net/bug.php?id=73099

[Nish Aravamudan (nacc)]

Thank you for doing that, I do believe it is an upstream issue
(possibly for all SAPIs) that php.ini parse failure doesn't
immediately lead to exit.