Re: About non-blocking mode usage

[Aris Adamantiadis <aris@xxxxxxxxxxxx>]

Hi,

This is exact to the precision that it's waiting for the socket to be
readable (and not writable). Since SSH has internal channel windowing,
we use a write buffering strategy to avoid blocking the app when a write
is pending.

Aris

Le 4/11/14 18:58, Doug Judd a écrit :
> I confronted this issue as well.  The documentation is lacking in this
> area.  It appears that most of the APIs, when they return SSH_AGAIN,
> it means that you need to wait until the underlying socket becomes
> ready for reading.  I found this to be true for both connection
> establishment APIs and command execution APIs.  I suppose that if you
> issue commands that are large, ssh_channel_request_exec() could return
> SSH_AGAIN and require you to wait until the socket becomes ready for
> writing, however, I'm not 100% sure on this.
>
> - Doug
>
> On Thu, Oct 30, 2014 at 4:55 PM, Liang Qin <lqin@xxxxxxxxxx
> <mailto:lqin@xxxxxxxxxx>> wrote:
>
>     Does anyone has an example of how to use the non-blocking mode for
>     libssh? 
>     This include ssh_connect, ssh_userauth functions, ssh_channel
>     functions etc. these functions will return again in non-blocking
>     mode, in the test examples we see it just keep trying until we get
>     ssh_ok, but this is not efficient and will cause high cpu usage, I
>     know we should use poll or ssh_poll, but I did not find any docs
>     or examples that shows how to use it.
>
>     Could anyone give me some example?
>     Thanks 
>
>
>
>
> -- 
> Doug Judd
> CEO, Hypertable Inc.