[RCU] Session timeouts all over the place

Thomas Bruederli thomas at roundcube.net
Wed May 2 10:05:44 CEST 2012


Michael Heydekamp wrote:
> [...]
> 
>> In order to take another attempt to track down the issues of session
>> timeouts, I suggest you to do the following:
>>
>> * Install a recent version of Roundcube (preferably 0.8-beta) from scratch.
> 
> That would be our third RC installation then. I'll have a word with my main
> admin, but wouldn't it be more helpful to stick with the current 0.7.2
> installation if we really want to track down this stuff? Because with this
> environment we are indeed seeing the problems.
> 
> If we would install 0.8-beta and the problems would seem (!) to be gone, we
> wouldn't still know why exactly. And we would never know if at some stage
> they wouldn't appear again. In some way I'm sort of "happy" to see those
> problems, as this is the only chance to track them down. Getting rid of
> bugs/problems by coincidence would be a somewhat, uh, sloppy way.

We already tried to track them down with your current version but obviously
without success. In the meantime we even made some changes to improve the
session validation and they might already have solved your issue as well.
Either you install the latest SVN in order to verify or we have to provide
patches which you can apply and test again. I just wanted to make sure that
you have a fresh installation without any hacking or patching being done so
far.
> 
>> * Use the default configuration values where possible
> 
> I'd say we have a pretty default configuration already (or still, as you
> want). Well, I've enabled QP encoding for instance (dunno why the default is
> disabled), set the line length to 76 instead of 72, disabled 'format=flowed'
> temporarily (I remember that it was called 'format=flawed' when I was still
> developing our own mail client years ago, but I can't recall the reasons in
> detail anymore), and we had to set 'imap_force_lsub' to true due to dovecot
> - all in all that's it, I'd say.
> 
>> * Do not install or activate any plugins
> 
> Hmm, uh, some of them I/we really need. But for testing purposes, I may do
> it if it helps.

I'm sure you need them but we're in the process of finding a bug. Therefore
we first have to locate it and this is the way to do so. I'm not saying
you'll never be able to use these plugins again.
> 
>> * Truncate database tables "session" and "cache"
> 
> This task I will have to assign to my main admin. I'm not touching this.
> 
>> * Start using that fresh installation of Roundcube with your IE8. If
>> you still experience session problems, try with other browsers such as
>> Firefox or Chrome.
> 
> Really...? If the problems (or a part of them) should be browser-related,
> how could we track them down then? RC should run with IE as well, shouldn't
> it?

Again, we're trying to track down the problems here and one big factor in a
web service is the browser. Sure, it has to work with IE8 or whatever
browser but with this, I'd like to learn whether there are differences in
session timeouts when using different browsers (from the same machine I
forgot to mention).
> 
> Plus that I'm reluctant to scatter my system with all sorts of software and
> browsers. I'm not a fan of IE (not that I'm being misunderstood in this
> respect), but it's the browser I'm using, not more not less.

That's certainly up to you. But if you want to help investigating this
problem, that's one thing I ask for. We cannot help you if everything is
working fine for us with other browsers or even with IE8 but yours has some
different settings or plugins installed which might cause some weird
behavior. I'm just trying to locate the component which causes your session
timeouts.
> 
>> * Enable the config option 'log_session' and check
>> <roundcubedir>/logs/session if you ever loose a session.
> 
> Done. But just a few minutes ago (and two hours too late), as I had a
> session timeout again two hours ago. Hopefully next time I'll be able to
> report something meaningful.

I'm sure there'll be another session timeout to be logged :-)

Thanks,
Thomas


More information about the users mailing list