Re: Asynchronous DNS

[Alain Schneble]

Eli Zaretskii <address@hidden> writes:

>> From: Alain Schneble <address@hidden>
>> CC: Eli Zaretskii <address@hidden>, <address@hidden>
>> Date: Sun, 14 Feb 2016 12:19:03 +0100
>> 
>> Lars Ingebrigtsen <address@hidden> writes:
>> 
>> > Alain Schneble <address@hidden> writes:
>> >
>> >> The following patch removes the separate dns_processes list in favour of
>> >> reusing Vprocess_alist.  The latter list is now used to loop over all
>> >> processes -- to check if pending DNS requests have completed.
>> >>
>> >> Do you agree whith this change?
>> >
>> > Well, it changes a test in the network loop that is basically a if (!
>> > NILP (Qnil)) in 99.9999% (plus some nines) of the cases with code that
>> > loops through all processes in the network loop all the time.
>> 
>> True.  Seems like I was too naive. I thought that these additional
>> cycles were negligible (even though it's O(n)...).
>
> We could traverse the list when Emacs is idle and maintain the result
> in a cache variable.

What exactly do you mean by "when Emacs is idle"?  The loop is in
wait_reading_process_output and I thought that this is "kind of"
executed when Emacs is idle.  Or do you refer to idle timers?  I tried
to analyze the call graph starting from command_loop.  I must be missing
something...  Is there yet another concept of "idleness" in Emacs?