Bug #47013 for POE-Component-Client-Keepalive: SF_QUEUE race condition (no test case)

Tue Jun 16 12:34:43 2009 loofort [...] gmail.com - Ticket created

Subject:	bug-POE-Component-Client-Keepalive
Date:	Tue, 16 Jun 2009 19:34:14 +0300
To:	bug-POE-Component-Client-Keepalive [...] rt.cpan.org
From:	Ilia Kandrashou <loofort [...] gmail.com>

When POE::Component::Client::Keepalive fires ka_request_timeout event for request with no wheel (request is in SF_QUEUE), then it executes respond_with_error, but doesn’t remove request from SF_QUEUE. So SF_QUEUE never become empty. And when new request executes _ka_add_to_queue, and check if SF_QUEUE is not empty then do nothing. And if currently no active connection than ka_wake_up never be executed. From this moment all new request via POE::Component::Client::HTTP returns timeouted response.

Sun Jul 26 03:11:15 2009 RCAPUTO [...] cpan.org - Correspondence added

On Tue Jun 16 12:34:43 2009, loofort@gmail.com wrote: Show quoted text

> When POE::Component::Client::Keepalive fires ka_request_timeout event > for request with no wheel (request is in SF_QUEUE), then it executes > respond_with_error, but doesn’t remove request from SF_QUEUE. > > So SF_QUEUE never become empty. > > And when new request executes _ka_add_to_queue, and check if SF_QUEUE > is not empty then do nothing. And if currently no active connection > than ka_wake_up never be executed. > From this moment all new request via POE::Component::Client::HTTP > returns timeouted response.

Hi! Do you have a test program I could run to trigger this? I'll try to make one myself, but it could be a while.

Sun Jul 26 03:11:16 2009 The RT System itself - Status changed from 'new' to 'open'

Thu Jul 30 06:04:41 2009 loofort [...] gmail.com - Correspondence added

Subject:	Re: [rt.cpan.org #47013] bug-POE-Component-Client-Keepalive
Date:	Thu, 30 Jul 2009 13:04:21 +0300
To:	bug-POE-Component-Client-Keepalive [...] rt.cpan.org
From:	Ilia Kandrashou <loofort [...] gmail.com>

Sorry, but a can't provide you with test prog. I discovered this bug in deeply integrated part of big program. 2009/7/26 Rocco Caputo via RT <bug-POE-Component-Client-Keepalive@rt.cpan.org>: Show quoted text

> <URL: https://rt.cpan.org/Ticket/Display.html?id=47013 > > > On Tue Jun 16 12:34:43 2009, loofort@gmail.com wrote:

>> When POE::Component::Client::Keepalive fires ka_request_timeout event >> for request with no wheel (request is in SF_QUEUE), then it executes >> respond_with_error, but doesn’t remove request from SF_QUEUE. >> >> So SF_QUEUE never become empty. >> >> And when new request executes _ka_add_to_queue, and check if SF_QUEUE >> is not empty then do nothing. And if currently no active connection >> than ka_wake_up never be executed. >> From this moment all new request via POE::Component::Client::HTTP >> returns timeouted response.

> > Hi! Do you have a test program I could run to trigger this? I'll try > to make one myself, but it could be a while. > >

Sun Feb 14 16:32:55 2010 RCAPUTO [...] cpan.org - Correspondence added

I haven't been able to reproduce the problem. I am leaving the ticket open and flagged as "stalled". Perhaps someone will contribute more information later.

Sun Feb 14 16:32:56 2010 RCAPUTO [...] cpan.org - Status changed from 'open' to 'stalled'

Wed Jan 26 16:49:51 2011 RCAPUTO [...] cpan.org - Subject changed from 'bug-POE-Component-Client-Keepalive' to 'SF_QUEUE race condition (no test case)'

Sun Mar 04 21:02:16 2012 RCAPUTO [...] cpan.org - Correspondence added

Nobody else has reproduced this in just over a year. I assume it was fixed as a side effect of some other commit, but I can't be sure. Please reopen with more information if the problem happens again. Thanks!

Sun Mar 04 21:02:18 2012 The RT System itself - Status changed from 'stalled' to 'open'

Sun Mar 04 21:02:19 2012 RCAPUTO [...] cpan.org - Status changed from 'open' to 'resolved'