[prev in list] [next in list] [prev in thread] [next in thread] 

List:       squid-dev
Subject:    Re: [PATCH 5/8] reconfiguration leaks: objects tied to http_port
From:       Amos Jeffries <squid3 () treenet ! co ! nz>
Date:       2014-07-14 9:55:29
Message-ID: 53C3A911.2050507 () treenet ! co ! nz
[Download RAW message or body]

On 13/07/2014 4:59 p.m., Amos Jeffries wrote:
> On 23/06/2014 12:30 a.m., Amos Jeffries wrote:
>> On 17/06/2014 9:15 a.m., Alex Rousskov wrote:
>>>
>>> PortCfgPointer is not a reference counting pointer.
>>
>> There is no remaining reason for that since we converted the TcpAcceptor
>> to emitting MasterXaction. The PortCfg pointer is not actually passed as
>> a parameter anywhere. Just one buggy piece of code which should have
>> been implemented differently.
>>
>>
>> <snip>
>>>
>>> Most likely, we should use the refcounting API for port pointers. Until
>>> that (or a better) solution is implemented, we should either
>>>
>>
>> The attached (rough) patch converts the PortCfgPointer to reference
>> counted and fixes all parsing errors resulting from the change. Most of
>> the issues were due to use of raw-pointers and explicit
>> cbdataReference*() API.
>>
>> Still have to add stubs to fix make check linkage errors and do some run
>> testing.
>>
> 
> If there are no objections I will commit the slightly more polished
> version of this patch shortly.
> 
> Amos
> 

Applied.

SSL state details may still be leaking, but I no longer see any issue
with SSL cleanup patches being applied so long as the relevant cleanup
is performed via the AnyP::PortCfg destructor.

Amos

[prev in list] [next in list] [prev in thread] [next in thread]