Phalcon segfault php-fpm

[tunhvn]

I'm using phalcon 1.3.2, php-fpm 5.5.16 (remi package), centos 6.5.
Nginx frequently return the error "502 bad gateway".

Nginx error log:
2014/09/12 11:57:19 [error] 26060#0: *82 recv() failed (104: Connection reset by peer) while reading response header from upstream, client: , server: , request: "POST HTTP/1.1", upstream: "fastcgi://127.0.0.1:9000", host: "", referrer: ""

Php-fpm error log:
[12-Sep-2014 11:57:19] WARNING: [pool www] child 26045 exited on signal 11 (SIGSEGV) after 82.764425 seconds from start
[12-Sep-2014 11:57:19] NOTICE: [pool www] child 26079 started

CentOS syslog:
Sep 12 11:57:19 CentOS kernel: php-fpm[26045]: segfault at 18 ip 000000000058d839 sp 00007fff9460e130 error 4 in php-fpm[400000+323000]

I think it's a bug in Phalcon 1.3.x core. Someone has the same issue #2516

Want to back this issue? Post a bounty on it! We accept bounties via Bountysource.

[duythien]

@tunh

Can you tun off xdebug, it working with me

[tunhvn]

I dont install Xdebug. It's exactly a bug of phalcon core :-(

[duythien]

I'm not sure . i used system very like you and that working

[tunhvn]

I think Phalcon has some conflict with zend_extension, ie opcode caching, but I can't disable opcache because of some heavy features in system, so disappointed

[Rewt0r]

Phalcon works fine for me with OPCode caching. I do notice a lot of weird behaviour if xdebug is enabled.

[sjinks]

@tunh do you happen to have the code that reproduces this issue?

[tunhvn]

@sjinks Most issue from ajax actions which return json. Now when I disable opcache, everything is okay

[brandonlamb]

There are at least two open issues for 1.3.x and opcache blowing up, FYI.

[sjinks]

Any chance anyone could generate a backtrace?

[tunhvn]

Have the same issue in Phalcon 1.3.3, dont have any fix?

[iansltx]

Was getting random segfaults all over the place as well, with 5.6.2, 1.3.3 and 1.6.4 versions of PHP, Phalcon and nginx, respectively. Was using Request::getHeaders() multiple times in a single request, and when i stopped doing that (caching the header list in user space instead) that resolved the issue. Crappy workaround, but things seem to be more stable now.

[ZeroxLim]

Ian suggestion to cache the result from Request::getHeaders() and return cache on subsequence call works.
Hopefully Phalcon team can look into it.

[andresgutierrez]

Could you please try again using Phalcon 2 or close this issue as 1.3.x is no longer supported

[Jurigag]

So Request::getHeaders() is causing this? Need to check this on 3.0.x.

[sergeyklay]

I'm closing this issue due to the lack of any reaction. I'll open it again if the need arises