Re: sftp_aio_wait_read() and SSH_AGAIN

[Edrusb <lib.ssh@xxxxxxxxxx>]

On 26/07/2025 06:03, Eshan Kelkar wrote:
> Hi Edrusb,

Hi Eshan,

Thank you for your quick and clear response!

[...]

> Or is there some reason (that I am overlooking) due to which this 
> could not be done in your case (or does not work ?) ?

Everything works fine, no worries! I was just not aware of the 
blocking/no-blocking mode of sftp_aio_wait_read() function, and wondered 
how to deal with possible situation of SSH_AGAIN returned code.

>     Or, asked another way: when sftp_aio_wait_read() returns
>     SSH_AGAIN, can
>     a program call it again right away as there is no chance it returns
>     again SSH_AGAIN and thus there is no need to wait/sleep for some time
>     between calls in that context, to avoid wasting CPU cycles and
>     creating
>     unnecessary load on the system?
>
>
> Yes, a program can call sftp_aio_wait_read() again right away after it 
> returns. However, when such blocking behaviour is needed, it is 
> recommended to use the aforementioned approach instead, i.e:
>
> 1. Open a sftp_file in blocking mode (default) or set to blocking mode 
> using sftp_file_set_blocking()
> 2. Call sftp_aio_wait_read() which should wait (or will report 
> failure) till expected data comes.

If the blocking mode is the default, that's fine, as I have not changed 
it that's what I'm using. And thus, I understand that in that mode I 
should not be worried by a returned code of SSH_AGAIN, perfect!

I would suggest adding a note in the sftp_aio_wait_*() documentation 
telling that SSH_AGAIN is only returned in non-blocking mode and add 
also a pointer to the sftp_file_set_blocking()/nonblocking() functions.  
My 2 cents ;)

> [...]
> Regards,
> Eshan Kelkar

Best Regards,
Edrusb