Re: non-blocking connect()

[Kim F. Storm]

Kazu Yamamoto (山本和彦) <address@hidden> writes:

> Hello,
>
>> > I'm not sure but I guess this patch does not solve the problem 
>> 
>> Can you try it, please.
>
> My friend is now using your patch. He has not met the problem so far.

So no definitive evidence yes, but it's at least an improvement...
At least it doesn't have any ill effects, so I'll install it.

>
> But I would point out a few things.
>
> (1) potential infinite loop.
>
>    If connect() is always interrupted, your code falls into infinite
>    loop. If you would like to stick to the current logic, I'd suggest
>    to prepare a limitation.

I don't know what would continue to interrupt the connect.  
There are other places in the code which would also loop in that case.

There is a QUIT in the loop, so the user can break out of it with C-g.

>
> (2) The block starting with
>       if (xerrno == EADDRINUSE && retry < 20)
>     is not necessary anymore. This is only effective in the original
>     code.
>

True.  Actually, it was disabled with #if 0 ... #endif in the patch.

-- 
Kim F. Storm  http://www.cua.dk