A message for data waiting at a socket?

[Doug Lerner]

On 1/4/04 1:53 AM, "Dar Scott" <dsc at swcp.com> wrote:

> 
> On Saturday, January 3, 2004, at 07:27 AM, Doug Lerner wrote:
> 
>> Never mind my last message - I did get it to "push" according to your
>> suggestion.
> 
> But I don't see how that would go into high-CPU consumption.

It didn't. It worked fine. That's why I said never mind my last message. :)

I guess the login in Rev for this is "when it successfully finishes a read
until CRLF send the message on, otherwise let other handlers continue
working and in the meantime nicely share cpu time while I am waiting for
more data."

At any rate, it works fine.

> 
> I usually don't use "until <lineEnd>", though I have no argument
> against it.  Maybe that changes the behavior.  I filter for messages at
> the receive handler.
> 
> I do get a high-CPU consumption state when the sender keeps sending
> gobs of data.  In a GUI app, this is hard to get out of since pending
> messages normally have higher priority over event messages.  I choke
> this two ways.  One is to put a delay between the read handler and the
> next read.  The other is to detect the situation and shut down the
> connection.  (I use other methods for accept.)

OK. A good caution to look out for.

doug

> 
> Revolution does "push" at the end of a write, but there might not be a
> way to insist on that kind of thing in other applications.
> 
> I'm curious about the high-CPU consumption.  If there are cases in
> which it can occur, I'd like to know about them.
> 
> Dar Scott
> 
> _______________________________________________
> use-revolution mailing list
> use-revolution at lists.runrev.com
> http://lists.runrev.com/mailman/listinfo/use-revolution