Re: Asynchronous DNS

[Alain Schneble]

Lars Ingebrigtsen <address@hidden> writes:

> Andreas Schwab <address@hidden> writes:
>
>> Alain Schneble <address@hidden> writes:
>>
>>> True, but my assumption was that it would be illegal to call
>>> `accept-process-output' say, for example, from within
>>> `set-process-coding-system' for the sake of waiting for the DNS resolve
>>> to complete as it might have unwanted side effects.
>>
>> You have to wait for the completion event in some way.
>
> Yeah, these blockers would basically have to be a version of this thing
> from `send_process':
>
>  while (! new_function_that_says_the_process_is_ready (proc)) 
>   wait_reading_process_output (0, 20 * 1000 * 1000, 0, 0, Qnil, NULL, 0);
>
> I think.  Which is what send_process does on EWOULDBLOCK.

Ok, now I see, thanks to both of you.  In the example above, wait_proc
is NULL.  Shall it be set to the process to wait for in a "real"
implementation?  Together with passing a sigevent struct to
getaddrinfo_a, so the process gets signaled?  Currently, this signaling
is not used, as it polls for completion.  I hope I'm not mixing things
up...