Re: Interesting APC behaviour

On 10/29/2012 06:57 AM, Gopal Vijayaraghavan wrote: >>> I've found that it behaves rather inconsistently when people start >>> returning values from scripts. >>> >>> <?php >>> function f($a = array(1)) { return $a; } >>> return f(); >>> >>> has been a suspect case for some segvs. >> >> Sorry, I didn't understand. Do you mean my change to include_once >> override is causing new segvs or are you saying in general our >> include_once override implementation has historically had problems? > > Yes, it has had troubles with people who return data from main scripts > with inclusion. I haven't seen that recently. If we have a reproducable case for this it doesn't sound like it would be hard to fix. >> I'd love to hear your suggestion for a simpler non-override way to solve >> this particular inconsistency. Clearly the way the included_files check >> is done currently is not ideal and I don't really see how we can solve >> this in APC without doing an override since the actual check is done in >> PHP's include_once code. > > I'd marked include_once_override for eventual deletion in my head, but > I will give this a bit more thought. > > I suspect that the apc file stack - APCG(cache_stack) might be a better > source for include once override instead of going & hitting the lock again > with a find(). But that would need a look-up optimized structure. Well, an optimization could be to re-use the cache_entry on the actual compile, but for now this implementation seems to be working quite well actually. It has the added benefit of removing the call to VCWD_REALPATH(). -Rasmus