Re: client_side and comm_close

Henrik Nordstrom wrote:
> lör 2008-04-19 klockan 17:24 +0300 skrev Tsantilas Christos:
>> If I am not wrong the bug which triggered this discussion is the bug2309
>> which is squid3.0 bug where AsyncCalls does not exists.
>
> Is it? I wasn't aware this was the problem we are discussing. 2309 is
> from what I can tell completely unrelated to comm_close or asyncness.
> Exactly what triggers it we don't know yet, other than for some reason
> two comm_read() calls gets scheduled on the same FD.
>

Ok, maybe the bug2309 is unrelated to the comm_close but it is related
to the client_side code complexity, and this is the first problem.....

>>From what I can tell 2309 is caused by the comm_read() refactoring,
> where a more low-level commSetSelect() API were used previously where it
> was no error to reset the event handler for an comm event while one was
> already set up on the fd..
>
> What triggered the comm_close discussion was the pconn closing issue
> from what I remember.
>

I am not saying that all are OK with the comm_close. Looks that there is
a problem with this function, and this is the second problem....

This why I am saying that there are two different problems here....
Received on Tue Apr 22 2008 - 13:45:43 MDT